write research paper chatbot

Have a language expert improve your writing

Check your paper for plagiarism in 10 minutes, generate your apa citations for free.

• Knowledge Base

Using AI tools

How to Write a Paper with ChatGPT | Tips & Examples

Published on August 17, 2023 by Koen Driessen . Revised on November 16, 2023.

A good research paper demonstrates your academic writing skills and knowledge of your chosen topic.

Your research paper should be based on in-depth independent research. However, generative AI tools like ChatGPT can be effectively used throughout the research process to:

• Brainstorm research questions
• Develop a methodology
• Create an outline
• Find sources
• Summarize and paraphrase text
• Provide feedback

Instantly correct all language mistakes in your text

Upload your document to correct all your mistakes in minutes

Table of contents

Brainstorming research questions, developing a methodology, creating an outline, finding sources, summarizing or paraphrasing text, receiving feedback, other interesting articles, frequently asked questions.

You can use ChatGPT to help formulate research questions , brainstorm potential thesis statements , or narrow down a broad subject area to a specific topic that interests you.

However, make sure that the outputs make sense and meet the requirements of your assignment. You can adapt these outputs manually, or use further prompts to ensure they meet your needs. For more information, see our tips for creating effective ChatGPT prompts .

• Generate three possible research questions for a research paper on the following topic: “The role of virtual reality in enhancing training and skill development in various industries.”
• “What are the cognitive and skill development benefits of virtual reality-based training compared to traditional training methods across different industries?”
• “How does the level of immersion in virtual reality training impact knowledge retention and skill transfer in fields such as medicine, aviation, and manufacturing?”
• “What are the key factors that influence the successful integration and adoption of virtual reality technology for training purposes in diverse industries, and what challenges hinder its widespread implementation?”

The only proofreading tool specialized in correcting academic writing - try for free!

The academic proofreading tool has been trained on 1000s of academic texts and by native English editors. Making it the most accurate and reliable proofreading tool for students.

Try for free

Your paper should clearly indicate the data collection and analysis methods you used in your research.

You can use ChatGPT to help decide what kind of research you will conduct (e.g., qualitative or quantitative ), how you will analyze the data, and how you will mitigate potential research bias .

Bear in mind that you should be able to rationalize your chosen methodology . For example, you should ensure that it is either a standard methodology in the field or can be justified.

• Suggest three quantitative research methods appropriate to the following question: “What are the cognitive and skill development benefits of virtual reality-based training compared to traditional training methods in the field of healthcare?”
• Experimental Study: Conduct a randomized controlled trial (RCT) where healthcare professionals are randomly assigned to either a virtual reality-based training group or a traditional training group. Both groups would receive training in specific healthcare tasks, and their cognitive and skill development outcomes would be measured and compared. Performance metrics, such as accuracy, speed, and efficiency, could be quantitatively assessed to determine the benefits of virtual reality-based training over traditional methods.
• Survey Research: Design a comprehensive survey targeting healthcare professionals who have undergone both virtual reality-based training and traditional training. The survey could include Likert-scale questions and structured response options to gauge their perceived cognitive and skill development benefits from each training method. Additionally, open-ended questions could be included to gather qualitative insights on specific strengths and weaknesses of each training approach.
• Task Performance Analysis: Implement a task-based assessment that simulates real-life healthcare scenarios, where participants are evaluated on their ability to handle critical situations. Participants would go through both virtual reality-based training and traditional training before being assessed in the simulated scenarios. Their performance in the assessment could be objectively measured using predefined criteria, allowing for a quantitative comparison of skill development between the two training methods.

A clear outline can help structure your paper, organize your arguments effectively, and make your research more efficient.

You can use generative AI tools like ChatGPT to formulate a potential outline to help guide your research. You can be as specific as you like, giving an overview of your research objectives , main arguments, and any evidence you intend to include in your paper.

However, ChatGPT outputs sometimes contain factual inaccuracies. It’s important to check the headings and subheadings suggested in the output and adapt them to meet your needs.

• Create an outline for an academic research paper with the following research question:“What are the cognitive and skill development benefits of virtual reality-based training compared to traditional training methods in the field of healthcare?”The paper will be about 10 pages long.

A. Theoretical basis for cognitive and skill development in training

B. Application of cognitive theories to virtual reality-based training

C. The role of immersive learning environments in skill acquisition

D. Conceptual framework for the study

IV. Research Methodology

A. Research design and approach (e.g., quasi-experimental, pre-post intervention)

B. Selection of participants and sample size calculation

C. Description of the virtual reality-based training and traditional training programs

D. Instruments and measures for assessing cognitive and skill development

E. Data collection procedures and ethical considerations

V. Data Analysis

A. Overview of statistical methods used for data analysis

B. Treatment of data and data cleaning

C. Quantitative analysis of cognitive and skill development outcomes

D. Presentation of findings with charts and tables

VI. Results

A. Presentation of the results of the cognitive and skill development assessments

B. Comparison of outcomes between virtual reality-based training and traditional training

C. Statistical significance and effect size analysis

D. Interpretation of the results and implications

VII. Discussion

A. Interpretation of findings in the context of the research question

B. Comparison of results with existing literature

C. Explanation of how virtual reality-based training impacts cognitive processes and skill development

D. Limitations of the study and areas for future research

VIII. Practical Implications

A. Practical applications of virtual reality-based training in healthcare settings

B. Recommendations for integrating virtual reality training in healthcare education

C. Benefits of virtual reality for continuous professional development in healthcare

IX. Conclusion

A. Summary of key findings

B. Reiteration of the research question and hypothesis

C. Contributions of the study to the field of healthcare training

D. Concluding remarks

X. References

You can use ChatGPT to help find sources relevant to your research. However, it’s important to be aware of the limitations of ChatGPT (for example, it occasionally cites nonexistent sources).

Instead of asking ChatGPT for specific source recommendations, try asking it for recommendations on the types of sources relevant to your research topic. You can also use it to generate a list of relevant keywords to use in your literature search.

• What types of sources could I use to write a research paper on the following research question?“What are the cognitive and skill development benefits of virtual reality-based training compared to traditional training methods in the field of healthcare?”
• Academic Journals: Peer-reviewed journals are valuable sources of scholarly research. Look for articles in medical education, cognitive psychology, virtual reality applications in healthcare, and technology-enhanced learning.
• Research Studies: Seek out published research studies that specifically investigate the impact of virtual reality-based training on cognitive and skill development in healthcare professionals.
• Systematic Reviews and Meta-Analyses: These studies provide comprehensive overviews of existing research on a specific topic, making them a valuable source for summarizing and analyzing the overall evidence.
• Books and Book Chapters: Academic books and book chapters authored by experts in the fields of medical education, virtual reality, cognitive science, or instructional design can provide in-depth insights and theoretical foundations.
• Conference Proceedings: Proceedings from relevant conferences and symposiums may present the latest findings and advancements in virtual reality-based training and healthcare education.
• Educational Institution Websites: Universities and medical schools often publish research papers, case studies, and whitepapers related to medical education and virtual reality training.
• Industry Reports: Reports and whitepapers from healthcare organizations or technology companies exploring the benefits of virtual reality-based training in the healthcare sector.
• Government and Non-profit Organizations: Look for reports and publications from government agencies and non-profit organizations that focus on medical education and healthcare training.
• Online Databases: Utilize academic databases such as PubMed, Google Scholar, IEEE Xplore, and ScienceDirect to search for peer-reviewed research articles and conference papers.
• Expert Interviews: If possible, conduct interviews with experts in medical education, cognitive science, and virtual reality to gain valuable insights and perspectives.
• Surveys and Polls: Refer to reputable surveys and polls related to virtual reality-based training in the healthcare industry for additional statistical data.

Check for common mistakes

Use the best grammar checker available to check for common mistakes in your text.

Fix mistakes for free

You can use ChatGPT to paraphrase or summarize text. This can help you to condense sources to their most important points and explore new ways of expressing your ideas.

Alternatively you can use the more specialized tools featured on Scribbr’s AI writing resources page (including Scribbr’s free text summarizer  and  Scribbr’s free paraphrasing tool ) which are designed specifically for these purposes and will give a smoother user experience.

• Provide a short summary of the following text:“One of the most profound applications of VR in healthcare is its use in surgical training. Traditionally, surgical training has relied on a combination of didactic lectures, observation of procedures, and hands-on experience in the operating room. However, these methods may not provide the ideal environment for novices to develop the necessary dexterity and spatial understanding essential for complex surgeries. VR overcomes these limitations by offering immersive, true-to-life simulations that allow surgeons to navigate through intricate anatomical structures, practice incisions, and learn to manage unexpected challenges that may arise during surgeries. This hands-on experience in a risk-free environment not only enhances technical skills but also builds confidence and reduces the learning curve for new surgical techniques.”
• Virtual Reality (VR) has a profound impact on healthcare, particularly in surgical training, as it provides immersive simulations that help surgeons develop dexterity, spatial understanding, and confidence for complex surgeries in a risk-free environment, ultimately enhancing their technical skills and reducing the learning curve.

When you’ve finished writing your research paper, you can use ChatGPT to receive feedback. You can be as specific as you like, selecting particular aspects the output should focus on (e.g., tone, clarity of structure, appropriateness of evidence to support your arguments).

You can also use ChatGPT to check grammar, spelling, and punctuation. However, it’s not designed for this purpose and occasionally misses errors. We recommend using a more specialized tool like Scribbr’s free grammar checker . Or, for more comprehensive feedback, Scribbr’s proofreading and editing service .

• Feedback the above text with regard to appropriateness of evidence and coherence of structure.
• The research paper demonstrates a well-structured and coherent presentation of the cognitive and skill development benefits of VR-based surgical training. The introduction effectively sets the context and rationale, while the literature review incorporates relevant historical and empirical evidence. The theoretical framework strengthens the foundation for the analysis.The methodology section could provide more details on the research design and study selection criteria. However, the sections on cognitive and skill development benefits present compelling evidence, supported by appropriate sub-sections.The exploration of improved patient outcomes lacks concrete examples but effectively outlines potential benefits. The section on challenges and future directions raises relevant points, but suggestions for solutions could enhance practicality.Overall, the paper combines theoretical, empirical, and practical evidence, offering a well-rounded approach. It establishes the significance of VR-based surgical training and concludes with a concise summary of key findings.

If you want more tips on using AI tools , understanding plagiarism , and citing sources , make sure to check out some of our other articles with explanations, examples, and formats.

• Citing ChatGPT
• Best grammar checker
• Best paraphrasing tool
• ChatGPT in your studies
• Is ChatGPT trustworthy?
• Types of plagiarism
• Self-plagiarism
• Avoiding plagiarism
• Academic integrity
• Best plagiarism checker

Citing sources

• Citation styles
• In-text citation
• Citation examples
• Annotated bibliography

Yes, you can use ChatGPT to summarize text . This can help you understand complex information more easily, summarize the central argument of your own paper, or clarify your research question.

You can also use Scribbr’s free text summarizer , which is designed specifically for this purpose.

Yes, you can use ChatGPT to paraphrase text to help you express your ideas more clearly, explore different ways of phrasing your arguments, and avoid repetition.

However, it’s not specifically designed for this purpose. We recommend using a specialized tool like Scribbr’s free paraphrasing tool , which will provide a smoother user experience.

No, it’s not a good idea to do so in general—first, because it’s normally considered plagiarism or academic dishonesty to represent someone else’s work as your own (even if that “someone” is an AI language model). Even if you cite ChatGPT , you’ll still be penalized unless this is specifically allowed by your university . Institutions may use AI detectors to enforce these rules.

Second, ChatGPT can recombine existing texts, but it cannot really generate new knowledge. And it lacks specialist knowledge of academic topics. Therefore, it is not possible to obtain original research results, and the text produced may contain factual errors.

However, you can usually still use ChatGPT for assignments in other ways, as a source of inspiration and feedback.

Cite this Scribbr article

If you want to cite this source, you can copy and paste the citation or click the “Cite this Scribbr article” button to automatically add the citation to our free Citation Generator.

Driessen, K. (2023, November 16). How to Write a Paper with ChatGPT | Tips & Examples. Scribbr. Retrieved April 15, 2024, from https://www.scribbr.com/ai-tools/chatgpt-research-paper/

Is this article helpful?

Koen Driessen

Other students also liked, how to write good chatgpt prompts, chatgpt citations | formats & examples, what are the limitations of chatgpt.

Koen Driessen (Scribbr Team)

Thanks for reading! Hope you found this article helpful. If anything is still unclear, or if you didn’t find what you were looking for here, leave a comment and we’ll see if we can help.

Still have questions?

"i thought ai proofreading was useless but..".

I've been using Scribbr for years now and I know it's a service that won't disappoint. It does a good job spotting mistakes”

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• 07 July 2023
• Correction 11 July 2023

Scientists used ChatGPT to generate an entire paper from scratch — but is it any good?

• Gemma Conroy

You can also search for this author in PubMed   Google Scholar

A pair of scientists has produced a research paper in less than an hour with the help of ChatGPT — a tool driven by artificial intelligence (AI) that can understand and generate human-like text. The article was fluent, insightful and presented in the expected structure for a scientific paper, but researchers say that there are many hurdles to overcome before the tool can be truly helpful.

Access options

Access Nature and 54 other Nature Portfolio journals

Get Nature+, our best-value online-access subscription

24,99 € / 30 days

cancel any time

Subscribe to this journal

Receive 51 print issues and online access

185,98 € per year

only 3,65 € per issue

Rent or buy this article

Prices vary by article type

Prices may be subject to local taxes which are calculated during checkout

Nature 619 , 443-444 (2023)

doi: https://doi.org/10.1038/d41586-023-02218-z

Updates & Corrections

Correction 11 July 2023 : An earlier version of this story implied that human researchers had guided ChatGPT through the steps to create a research paper. In fact, the data-to-paper tool developed by Kishony and Ifargan acted as an intermediary between ChatGPT and the researchers. The text has been corrected in several places to reflect this.

Reprints and permissions

Structure peer review to make it more robust

World View 16 APR 24

US COVID-origins hearing puts scientific journals in the hot seat

News 16 APR 24

Is ChatGPT corrupting peer review? Telltale words hint at AI use

News 10 APR 24

Researchers want a ‘nutrition label’ for academic-paper facts

Nature Index 17 APR 24

Rwanda 30 years on: understanding the horror of genocide

Editorial 09 APR 24

Three ways ChatGPT helps me in my academic writing

Career Column 08 APR 24

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

• Perspective
• Open access
• Published: 25 February 2023

Can artificial intelligence help for scientific writing?

• Michele Salvagno 1 ,
• Fabio Silvio Taccone 1 &
• Alberto Giovanni Gerli 2  

Critical Care volume  27 , Article number:  75 ( 2023 ) Cite this article

86k Accesses

220 Citations

379 Altmetric

Metrics details

A Correspondence to this article was published on 10 May 2023

A Comment to this article was published on 18 April 2023

A Correspondence to this article was published on 21 March 2023

A Correction to this article was published on 08 March 2023

This article has been updated

This paper discusses the use of Artificial Intelligence Chatbot in scientific writing. ChatGPT is a type of chatbot, developed by OpenAI, that uses the Generative Pre-trained Transformer (GPT) language model to understand and respond to natural language inputs. AI chatbot and ChatGPT in particular appear to be useful tools in scientific writing, assisting researchers and scientists in organizing material, generating an initial draft and/or in proofreading. There is no publication in the field of critical care medicine prepared using this approach; however, this will be a possibility in the next future. ChatGPT work should not be used as a replacement for human judgment and the output should always be reviewed by experts before being used in any critical decision-making or application. Moreover, several ethical issues arise about using these tools, such as the risk of plagiarism and inaccuracies, as well as a potential imbalance in its accessibility between high- and low-income countries, if the software becomes paying. For this reason, a consensus on how to regulate the use of chatbots in scientific writing will soon be required.

Introduction

A chatbot is an electronic system (generally a software) that simulates conversations by responding to keywords or phrases it recognizes and that can be integrated into various platforms, such as websites, mobile apps, and messaging platforms.

The Chatbot Generative Pre-trained Transformer (ChatGPT), developed by OpenAI, is a type of Artificial Intelligence (AI) software designed to simulate conversations with human users. This chatbot works through algorithms programmed to understand natural language inputs and answer with appropriate responses, either pre-written or newly generated by the AI. ChatGPT is constantly improved with reinforcement techniques, natural language processing, and machine learning, to improve its ability to understand and thoroughly respond to users’ needs. Concretely, you can conversationally ask anything and receive a fast and adequate human-like written reply to your questions or requests such as: (a) write a small text on a given topic; (b) get information on a topic of interest; (c) compose an email or message with a certain tone, specific content, and intended for a particular person; (d) correct the shape of a text or change its wording; (e) solve problems.

As such, this chatbot could also be used in scientific writing [ 1 ]. Indeed, ChatGPT could become a promising and powerful tool for tasks such as automated draft generation, article summarizing, and language translation, which may be useful in academic activities to make writing work faster and easier. However, the use of this tool in scientific writing raises some ethical concerns and should therefore be regulated.

ChatGPT in scientific writing

ChatGPT is already able to help medical researchers and scientists to write articles and abstracts, in literature research, to summarize data or information, to provide suggestions for structure, references, and titles, in language reviews to make the text more readable, or even to generate a full draft of a paper [ 2 ]. At this stage, although feasible, no article dealing with critical care medicine has been published so far using this application. Moreover, whether writing a manuscript dealing with a topic of critical care medicine would be easier or more difficult when compared to other medical disciplines remains unknown. Which are the limits of possible applications of these tools have to be clarified.

ChatGPT can generate text on a wide variety of topics. Recently, an entire letter to the editor was written by ChatGPT, answering the author's questions [ 1 ]. Artificial Intelligence cannot generate new ideas, but it can organize and develop the ones of the researcher, creating an initial draft. However, this appears to be a starting point for a human-based development of the text, as the automatic text generated is far from being a substitute for the knowledge, creativity, and critical thinking of human experts.

For the literature search, ChatGPT and AI research assistants (such as “elicit.org”) can aid researchers in the review process by finding academic papers, summarizing their conclusions, and highlighting areas of uncertainty (i.e., summarizing the recent randomized clinical trial on the use of extracorporeal membrane oxygenation for refractory cardiac arrest [ 3 ], Fig.  1 ). This could help physicians, as an example, to quickly gain an understanding of the current state of knowledge on a particular topic, and to identify potential gaps that need to be addressed, although the provided summary might be quite generic and do not critically analyze differences among studies.

Example of a summary of the existing literature on a specific topic

During the writing process, ChatGPT can help to generate an initial draft of a scientific paper and even suggest titles. By providing raw information, ChatGPT can also assist in composing the section on the methods used in the study, justify sample size and describe data analysis techniques. From author’s experience, when the manuscript has been finalized, ChatGPT is extremely effective for the editing process; formatting and language editing, rewriting a particularly complex sentence in a clearer way, and even summarizing the entire text in order to compose a suitable abstract, are feasible using this approach, although the results are not always satisfactory, but they certainly save time. ChatGPT has been recently asked to prepare regular discharge summaries [ 4 ], but the quality of the content has not been specifically evaluated. To our knowledge, a structured evaluation to determine the quality of the output has not been performed yet. More complicated writing processes, such as systematic review and meta-analyses, require the human intervention and ChatGPT might be helpful mainly for editing.

A future potential application of AI is in the automatic generation of figures, tables, and other visual elements of the manuscript, which can aid in summarizing data. These elements are important for the clarity and understanding of the manuscript, but they are often time-consuming to create.

Importantly, the writing process of a scientific paper requires, for the moment, the guidance and supervision of human researchers who are experts in the field to ensure the accuracy, coherence, and credibility of the content before being used or submitted for publication. Chatbots can help but needs the researcher’s input, and inadequate inputs would lead to inadequate results. For this reason, chatbots and AI, in general, should not replace human researchers' expertise, judgment, personality, and—in the end—responsibility.

Chatbots vs. human beings

As AI, ChatGPT has a superior advantage in terms of fast comprehending information deeply and connecting evidence to reach conclusions, compared to humans who have limitations in their ability to read a comprehensive range of literature and distinguish the connections between seemingly separate pieces of information.

Moreover, it may be difficult to recognize whether a paper is written by a chatbot or a human being [ 5 ], because chatbots use advanced techniques, such as natural language processing (NLP) and machine learning, to generate text that is similar to human writing. To detect the author is a complex task and requires thorough critical reading to reach a conclusion. However, a few characteristics might reveal that a paper was written by a chatbot, such as the lack of nuance, style, or originality, which could allow the identification by AI output detectors and skeptical human reviewers [ 6 ]. Interestingly, the same writing traits could be found in texts written in a language that is not an individual's native tongue. By focusing on these traits, it is possible that AI plagiarism detectors would identify non-native English language papers as AI-generated text. It would be interesting to find these tools' sensibility in detecting the authors of texts written by these two groups.

Nevertheless, the chatbot-generated text might lack the subtle phrasing and word choice that a human author might use to convey a certain meaning or tone. It might also be vaguer and contain inconsistencies that would not be present in a human-written paper. Instead, if the paper contains a high degree of structural and grammatical errors, it may suggest that it was written by a human being (but it should not be the other way around). Lastly, if the paper is discussing a topic that is very specific and highly technical, it is less likely that a chatbot could generate such a text, as it would require a deep understanding of the topic and the ability to generate scientific analysis and conclusions.

Ethical considerations

Ethical concerns could also limit the use of these chatbots for scientific writing [ 7 ].

The process of acquiring knowledge from others and writing new or review articles involves humans incorporating both what they have learned from others and their ideas. It is natural for humans to repeat the findings, statements, and written works of others, and thus to come close to committing plagiarism by presenting an idea without proper referencing to the original authors. AI or ChatGPT systems can commit plagiarism under this definition but can also be programmed to avoid copying others by rephrasing their work in a manner similar to what human authors do. However, using programs to reformulate sentences and writing to reduce the percentage of plagiarism (i.e., asking the software to rewrite a section written by other authors with different words) could not be considered acceptable in scientific research. If we define “plagiarism” as a mere act to copy someone else work, just rephrasing what it was written, regardless of the method used, and without adding anything personal, it is a violation of academic integrity. For this reason, journal editors should use programs to detect written content using AI to detect plagiarism better.

Second, the lack of an expert and critical human mind behind scientific work (which is the basis of the scientific method) could lead to a risk of perpetuating or amplifying existing biases and inaccuracies in the data, providing unfair results and hampering scientific growth. Whatever the use of AI, we believe that the presence of an expert in the field in conducting scientific activity and writing is a necessary cornerstone even to guarantee the quality of the work.

Third, the incredible development of AI tools can lead to a significant increase in publication numbers from some researchers, but not accompanied by a real increase in her or his experience in that field. Ethical issues can therefore arise regarding hiring professionals by academic institutions that score on the number of publications rather than on their quality.

Fourth, whether the ChatGPT should be mentioned within the authors of the manuscript written using this approach remains not adequately defined. Finally, if at the moment ChatGPT and other chatbot services are free of charge, it is not guaranteed that they will not become paying in the future. The introduction of charges to access these chatbots could lead to a further disparity between high- and low-income countries (as well as between more junior to older professionals), for scientific production, resulting in unfair facilitation for the formers with unpredictable consequences.

Chatbot as a tool in the ICU

In addition to scientific writing, ChatGPT has the potential to assist physicians in their hospital work by saving time and allowing them to focus on providing patient care. In an Intensive Care Unit (ICU), where constant monitoring of multiple patient information is required, such as treatment progression, laboratory values, microbiological results, and fluid balance calculation, ChatGPT can assist in several ways.

First, ChatGPT can provide general information about recognized ICU protocols: when given a specific request, ChatGPT would be able to generate a response (i.e., initial management of sepsis) by analyzing the input and identifying patterns in the text data that it has been trained on. Furthermore, ChatGPT has a distinct advantage over humans in quickly gathering, comprehending, and utilizing information. In future, it may eventually be taught to apply this knowledge in clinical practice by presenting the latest evidence to healthcare professionals, if not directly creating patient-tailored protocols to reflect new findings.

Currently, it could already be used for generating clinical notes by just providing raw information, such as daily summaries and discharge summaries, which could save time and increase accuracy. Moreover, it could help in communicating with patients and families, translating into other languages, and generating personalized messages with accurate and timely information about the patient's condition, which would not have the purpose of replacing, but add to, the doctor-patient (or doctor-relatives) relationship. Unfortunately, depending on the applicable legislation, privacy issues may arise when providing patients' information.

Anecdotally, when provided with clinical information, laboratory values, past medical history, and other relevant data, ChatGPT may already assist physicians in, for example, determining appropriate treatment options; however, it is important to note that the answers provided by the chatbot can be erroneous or dated, as the technology is still in development and may not have the level of accuracy, current knowledge, or precision that is desired. Nevertheless, it is believable that AI will continue to improve in the future and potentially provide more accurate and reliable results.

As next future perspectives, AI in general could potentially be trained to automatically extract and understand all the relevant information from electronic health records and by analyzing patient data, such as vital signs, laboratory results, and medical history, helping doctors to quickly access patient information, providing recommendations for interventions, and allowing more fast and comprehensive decisions. The implementations of these AI systems, which already exist in several areas of the world, with chatbots, could create an interacting system which would provide immediately elaborated information to the treating physician.

Conclusions

ChatGPT is an AI software potentially able to assist in the writing process of a scientific paper and can help in the literature review, identify research questions, provide an overview of the current state of the field, and assist with tasks, such as formatting and language review. Moreover, it can have applications in clinical practice as time-savers. As chatbot tools will be widely adopted in the near future, it is crucial to have international academic regulations in place to regulate their use in scientific writing and establish mechanisms for identifying and penalizing unethical usage. Chatbots are simply tools, they can assist human researchers but should not be used as a replacement for human researchers' expertise, judgment, and personality.

We have requested to ChatGPT “Could you review this paper and tell us if you (as an author!) agree to submit it to Critical Care Journal?”. The (non-modified) response from ChatGPT has been: “ As an AI model, I am not able to review or submit papers to journals as I am not a researcher or author. However, I can give you some feedback on the paper you've provided. The paper presents a clear and comprehensive overview of the potential uses of ChatGPT in scientific writing, specifically in the field of critical care medicine. The introduction provides a good background on chatbots and ChatGPT and how it can be used in various tasks. The paper goes on to discuss the potential uses of ChatGPT in scientific writing, such as draft generation, literature research and language review. The paper also touches on the ethical concerns and limitations of using ChatGPT in scientific writing. I would recommend that you review the paper again, check for grammar and spelling errors, and make sure that the citations and references are correctly formatted. It would be also important to have a proper conclusion that summarizes the main ideas of the paper and the future perspectives. It's important to keep in mind that ChatGPT should always be used in combination with the expertise and judgement of human experts and its output should be validated before it is used in clinical practice.”

Availability of data and materials

Not applicable.

Change history

08 march 2023.

A Correction to this paper has been published: https://doi.org/10.1186/s13054-023-04390-0

King MR. The future of AI in medicine: a perspective from a Chatbot. Ann Biomed Eng. 2022;51:291–5. https://doi.org/10.1007/s10439-022-03121-w .

Article   PubMed   Google Scholar  

Hutson M. Could AI help you to write your next paper? Nature Nature Research. 2022;611:192–3.

Article   CAS   Google Scholar  

Suverein MM, Delnoij TSR, Lorusso R, Bruinsma GJBB, Otterspoor L, Kraemer CVE, et al. Early extracorporeal CPR for refractory out-of-hospital cardiac arrest. New Engl J Med. 2023;388:299–309. https://doi.org/10.1056/NEJMoa2204511 .

Patel SB, Lam K. ChatGPT: the future of discharge summaries? Lancet Digit Health. 2023. https://doi.org/10.1016/S2589-7500(23)00021-3 .

Else H. Abstracts written by ChatGPT fool scientists. Nature. 2023;613:423–423.

Article   CAS   PubMed   Google Scholar  

Gao CA, Howard FM, Markov NS, Dyer EC, Ramesh S, Luo Y, et al. Comparing scientific abstracts generated by ChatGPT to original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers. BioRxiv. 2022. https://doi.org/10.1101/2022.12.23.521610v1 .

Article   PubMed   PubMed Central   Google Scholar  

Hammad M. The impact of artificial intelligence (AI) Programs on writing scientific research. Ann Biomed Eng. 2023;51:459–60.

Download references

Acknowledgements

For the writing of this article, we have not received funds or support from OpenAI, which is not associated with the entire process that led to the preparation of this article. The text, written with the support of the ChatGPT by OpenAI, has however been modified by the human authors, who assume full responsibility for form and content.

This research received no external funding.

Author information

Authors and affiliations.

Department of Intensive Care, Erasme Hospital, Université Libre de Bruxelles, 1070, Brussels, Belgium

Michele Salvagno & Fabio Silvio Taccone

Department of Clinical Sciences and Community Health, Università Degli Studi di Milano, 20122, Milan, Italy

Alberto Giovanni Gerli

You can also search for this author in PubMed   Google Scholar

Contributions

MS, FST, AGG conceived of the study, and participated in its design and coordination and helped to draft the manuscript; MS, FST, AGG participated in the design of the study; MS, FST, AGG wrote, reviewed and made extensive editing of the final content; FST conceived and took the screenshot. All authors have read and agreed to the published version of the manuscript.

Corresponding author

Correspondence to Michele Salvagno .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests, additional information, publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The original online version of this article was revised: shortly after initial publication of this article, the authorship and Acknowledgements were updated in line with Springer Nature authorship policies. Large Language Models (LLM), such as ChatGPT, do not currently satisfy our authorship criteria. An attribution of authorship carries with it accountability for the work, which cannot be effectively applied to LLMs.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Salvagno, M., Taccone, F.S. & Gerli, A.G. Can artificial intelligence help for scientific writing?. Crit Care 27 , 75 (2023). https://doi.org/10.1186/s13054-023-04380-2

Download citation

Received : 31 January 2023

Accepted : 21 February 2023

Published : 25 February 2023

DOI : https://doi.org/10.1186/s13054-023-04380-2

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Artificial intelligence
• Scientific writing
• Machine learning

Critical Care

ISSN: 1364-8535

• Submission enquiries: [email protected]

Exploring agent-based chatbots: a systematic literature review

• Original Research
• Open access
• Published: 01 June 2023
• Volume 14 , pages 11207–11226, ( 2023 )

Cite this article

You have full access to this open access article

• Davide Calvaresi   ORCID: orcid.org/0000-0001-9816-7439 1 ,
• Stefan Eggenschwiler 1 ,
• Yazan Mualla 2 ,
• Michael Schumacher 1 &
• Jean-Paul Calbimonte 1 , 3  

2628 Accesses

Explore all metrics

In the last decade, conversational agents have been developed and adopted in several application domains, including education, healthcare, finance, and tourism. Nevertheless, chatbots still need to address several limitations and challenges, especially regarding personalization, limited knowledge-sharing capabilities, multi-domain campaign support, real-time monitoring, or integration of chatbot communities. To cope with these limitations, many approaches based on multi-agent systems models and technologies have been proposed in the literature, opening new research directions in this context. To better understand the current panorama of the different chatbot technology solutions employing agent-based methods, this Systematic Literature Review investigates the different application domains, end-users, requirements, objectives, technology readiness levels, designs, strengths, limitations, and future challenges of the solutions found in this scope. The results of this review are intended to provide researchers, software engineers, and innovators with a complete overview of the current state of the art and a discussion of the open challenges.

Similar content being viewed by others

Examining Science Education in ChatGPT: An Exploratory Study of Generative Artificial Intelligence

Grant Cooper

AI-based chatbots in customer service and their effects on user compliance

Martin Adam, Michael Wessel & Alexander Benlian

An Overview of Chatbot Technology

Avoid common mistakes on your manuscript.

1 Introduction

Conversational agents have been proposed and designed to enable seamless interactions with people, through computer-based means for communication, language processing, interpretation, and dialogue exchange (Adamopoulou and Moussiades 2020 ). These agents have substantially evolved from its first incarnation, the seminal project ELIZA developed by Joseph Weizenbaum (Weizenbaum 1966 ). Ever since, conversational agents have leveraged on Natural Language Processing (NLP), state machine engines, and pattern matching with the intent of engaging in purposeful conversations with human users. Several milestones marked the technological evolution of conversational bots. Towards the end of the 1980 s, Rollo Carpenter developed Jabberwacky (Rollo 1997 ), a self-learning agent mainly employing contextual pattern matching to identify the best answer (accessible over the internet only later in 1997). Later, in 1994, the term ChatterBot made its first appearance, used by Michael Mauldin to describe conversational programs (Mauldin 1994 ). Nowadays, this term has been shortened to chatbot , and it is used on a daily basis to describe these technologies. In the 1990 s, considerable progress was made on conversational agent technologies, based on advances in Artificial Intelligence. For example, Richard Wallace developed ALICE (Artificial Linguistic Internet Computer Entity), which leveraged on heuristical pattern matching. Footnote 1

In the 2010 s, chatbot technologies started to gain adoption, outside the academic sphere, in industrial and mainstream applications. Apple was among the first to commercialize a personal assistant with conversational capabilities in 2011 with the release of Siri. Footnote 2 Initially based on the Active platform (Guzzoni 2008 ), it assisted iPhone users recognizing both written and spoken language. Other major technology companies released their virtual assistants shortly. Google Now for Android and iOS devices appeared in 2012, Footnote 3 evolving from a simple recommendation engine to a personal assistant able to dialogue with the user (similar to Siri). Footnote 4 Microsoft followed with Cortana, Footnote 5 which was released in 2014. The same year, Alexa was launched by Amazon, primarily targeting home automation and online shopping. Although it is not linked to any OS, it quickly gained adoption in the market (Etherington 2014 ). The widespread acceptance of these major companies’ virtual assistants and their usage of asynchronous text-based interactions stimulated instant messenger applications to release APIs for third-party development of chatbots (i.e., Telegram, Facebook Messenger, and WhatsApp), in addition to those mainly dedicated to customer services through their web pages.

The increasing adoption of chatbots has been boosted by anywhere/anytime availability, immediate response, confidentiality, social acceptance, and massive scalability. Leveraging on these aspects, chatbots have proven to be effective in a wide range of domains such as eCommerce (Cui et al. 2017 ), education (Winkler and Söllner 2018 ), and in particular for motivational (e.g., social network campaigns (Calvaresi et al. 2019 )) and support (e.g., customer management (Xu et al. 2017 ), eHealth (Calbimonte et al. 2019 ), and assisted-living scenarios (Fadhil and Gabrielli 2017 )).

Recent remarkable technological advancements are pushing the evolution of chatbots using keyword-based text recognition or static finite state machines (FSM) to interpret and orchestrate user interactions (today still representing a significant share of the market), to hybrid solutions merging NLP (for text recognition) and FSM (for the management of intentions and user stories) (DeepLink 2022 ). However, the solely FSM-based solutions still expose significant limitations, such as inadequate personalization, lack of real-time monitoring, reporting and customization, lack of mechanisms to integrate communities of chatbots, limited knowledge sharing capabilities, and the impossibility of deploying multi-domain campaigns within the same framework. These limitations are linked to the predominantly rigid architectures proposed in most existing approaches. These rely on very specific scenarios translated into chatbot logic, which have to be reprogrammed every time a new scenario arrives. This raises the costs of modifying the behavior of a chatbot and prevents administrators from adapting it to specific situations. Moreover, most chatbot solutions rely on monolithic and centralized data management strategies, making it hard to comply with privacy regulations (e.g., European Union’s General Data Protection Regulation – GDPR (Voigt and Von dem Bussche 2017 )). The sensitive nature of data collected through chatbot interactions makes it necessary to shift the control of personal data towards the users themselves, empowering them in the process. Many chatbot systems have used AI to boost the accuracy and user-experience of its interactions. Examples include the use of NLP to generate asynchronous follow-up questions (Rao et al. 2021 ), or the application of neural networks to perform emotion detection in chatbot conversations (Huddar et al. 2021 ). However, these AI techniques focus more on the generation of responses and monitoring conversational context, without considering the autonomous, decentralized and collaborative nature of chatbots.

Nevertheless, in the last decade, the trend of combining chatbots with multi-agent systems (MAS) models and technologies tried to mitigate the limitations mentioned above. Particular emphasis is given to application domains where the social and collaborative dimensions (e.g., crowd-sourcing, user profiling and personalization) is essential in the interaction with users. These features are particularly relevant for domains such as healthcare fostering behavioral change (Pereira and Díaz 2019 ), where the majority of the studies/contributions bridging chatbots and MAS can be found (Calbimonte et al. 2019 ; Calvaresi et al. 2019 ).

To better understand the current panorama of the different chatbot technology solutions employing agent-based approaches, this work presents a Systematic Literature Review (SLR) investigating application domains, end-users, requirements, objectives, technology readiness level (TRL) (European Commission 2017 ), designs, strengths, limitations, and future challenges of the solutions found in the literature. The goal is to provide a tool for researchers, software engineers, innovation managers, and other practitioners to investigate the current state of the art and discuss the open challenges.

The rest of the paper is structured as follows: Sect.  2 presents the methodology applied for performing the SLR. Section  3 presents the review planning phase, including the definition of the protocol and the research questions. Section  4 describes how the review was performed. Section  5 analyses the outcomes of the applied methodology structured according to the research questions. Section  6 discusses the obtained results, projecting them into the stated (by the primary studies) and envisioned (by the authors of this paper) future directions. Finally, Sect.  7 concludes the paper.

2 Systematic literature review methodology

The approach employed in this paper aims at being both rigorous and reproducible. It relies on the methodology outlined by Kitchenham (Kitchenham et al. 2009 ), which has also been employed in a similar contexts (Palmarini et al. 2018 ; Calvaresi et al. 2021b ; Anjomshoae et al. 2019 ; Mualla et al. 2019 ; Calvaresi et al. 2018 ). Figure  1 proposes a schematic representation of the adopted procedure. In particular, it comprises three stages:

Planning the review. This phase consists of defining the main generic question(s) and deriving Structured Research Questions , characterizing the entire search protocol, matching the requirements ( rigorousness and reproducibility ), and validating the protocol.

Performing the review. Entails the execution of the following planned activities: collection and selection of literature, literature elaboration, and disagreement resolution.

Dissemination. Includes analysis, documentation, reporting, and summary of the learned lessons.

Systematic literature review phases (Kitchenham et al. 2009 )

3 Review planning

This section describes the definition of the structured research questions and the development of the review protocol describing the search strategy , the inclusion and exclusion criteria , the biases and disagreement resolution , and the quality criteria .

3.1 Research questions

As introduced in Sect.  1 , the research community has proposed the usage of multi-agent-based chatbots in recent years, for different domains, stakeholders, and purposes. Therefore, the main research question can be contextualized in these terms as follows: How are agent-based chatbots characterized, envisioned, and employed? To better investigate such a question, we comply with the Goal-Question-Metric (GQM) approach introduced by (Galster et al. 2014 ; Kitchenham et al. 2010 ). Such an approach has been employed in several other studies in the computer science-related domain (e.g., augmented reality for maintenance (Palmarini et al. 2018 ), virtual reality for education (Radianti et al. 2020 ), explainable agents and robots (Anjomshoae et al. 2019 ), agents and blockchains (Calvaresi et al. 2018 )) and other domains (e.g., tourism (Yang et al. 2017 ; Calvaresi et al. 2021b ). The dimensions targeted in this study apply to “intelligent” technologies and research. In particular, they are scientific interest over the years, application domains, stakeholders, requirements, goals, technologies, advantages, limitations, countermeasures, and future research . By formulating questions addressing such aspects, provide investigations and analysis in support of practitioners (providing an aggregated understanding of the current works), new tech pioneers (understanding what has been tried and what might be future targets), and industrial researchers (to bring research ideas onto the real-world market). Thus, we devised a set of ten structured research questions.

To establish an understanding of the demographic evolution of agent-based chatbots, we inquire: How are the research efforts temporally and geographically distributed?

To elicit the domains on which the agent-based chatbots research focuses, we inquire: Which application domains have employed agent-based chatbots?

To clarify who are the stakeholders of agent-based chatbots, we inquire: Who are the users of the chatbot systems relying on the agent paradigm?

To formalize the requirements arranged w.r.t. the given stakeholders, we inquire: What are the requirements standing behind the employment of agent-based chatbots?

To explore what research tried to achieve with agent-based chatbots, we inquire: What are the objectives set for agent-based chatbots?

To better understand the technological characterization, we have structured SRQ6 in four sub-questions:

Which chatbot design (e.g., paradigms) and implementations have been proposed?

Which technologies have been employed in the proposed solutions?

Which technologies have been previously employed?

What is the Technology Readiness Level  (European Commission 2017 ) of the solutions proposed in the primary studies?

To explore the benefit of existing solutions, we inquire: What are the strengths of employing agent-based chatbots?

To identify the shortcomings of the existing solutions, we inquire: What are the limitations of employing agent-based chatbots?

To understand the measures employed by the authors to achieve their objectives and overcome the limitations, we inquire: What are the proposed solutions for the limitations identified in SQR8?

Finally, to foster the establishment of future objectives, we inquire: What are the future challenges for chatbot-based solutions envisioned by the primary studies?

3.2 Review protocol

The search strategy included the selection of the following information sources: IEEE Xplore, Footnote 6 ScienceDirect, Footnote 7 ACM Digital Library, Footnote 8 Citeseerx, Footnote 9 and Pubmed. Footnote 10 The selection of the keywords relied on the reviewers’ background and knowledge in the context of agent-based chatbots, and they include the following: Multi-agent system, MAS, agent-based, chatbot, conversational agent, virtual assistant, personal assistant. To increase the results’ accuracy, some keywords have been aggregated. For example, MAS was expanded to three different queries: (i) MAS + chatbot + virtual assistant, (ii) MAS + chatbot + personal assistant, and (iii) MAS + chatbot + conversational agent.

Each search query produced a set of articles added to the list of papers to be considered. The result of each query has been screened by the reviewers to evaluate the articles’ coherence with the study. In particular, title and abstract have been pre-processed according to the criteria presented in the next section.

3.2.1 Inclusion and exclusion criteria

The initial search collected 108 papers, hereafter referred to as primary studies . Additional filtering criteria have been applied (see Table  1 ). In particular, such criteria have been selected aiming at (i) avoiding multiple papers (usually incremental) describing the same work, (ii) bounding the time window for the investigation (e.g., excluding too old and less-relevant works, given the technological advancements), (iii) selecting works contributing to the actual investigated topic, and (iv) ensuring the presence of a tangible theoretical/practical contribution – avoiding purely visionary and blue sky studies). The criteria definition is usually quite specific per topic/review. Nevertheless, several studies recall similar criteria selections (Yang et al. 2017 ; Anjomshoae et al. 2019 ). Applying the criteria defined in Table  1 , we purged unrelated papers and narrowed them down to a set of 38 contributions. Three reviewers were instructed to verify the compliance of the papers with the aforementioned inclusion criteria. Each reviewer operated independently while filtering out the list of papers. After the filter process ended, a review process was established so that a paper was included if at least two reviewers agreed on it.

3.2.2 Biases and disagreement resolution policy

The policy for biases and disagreement resolution allows the reviewers to cross-examine each task to limit biases and resolve disagreements among themselves. In particular, during the articles selection task, three reviewers cross-validated the inclusion/exclusion. During the elaboration of the articles, uncertainties have been discussed during periodic meetings.

3.2.3 Features and quality criteria

Assessing the quality of the extracted information is crucial. The following set of features has been chosen to answer the structured research questions: Publication year, geographical localization, main purpose, context, kind of users involved, scenarios, level of abstraction \(\dagger\) , architectures and designs, development methodologies, techniques, technologies and devices, user needs coverage \(\ddagger\) , need - offered support relation, kind of disease or difficulties supported \(\ddagger\) , awareness provided, architectural evidence \(\ddagger\) , technological evidence \(\ddagger\) , technical evidence \(\ddagger\) , architectural limitations \(\ddagger\) , technological limitations \(\ddagger\) , technical limitations \(\ddagger\) , identified future directions, identified future challenges. The features annotated with ( \(\dagger\) ) are classified with C, P, or T as possible values, that respectively stand for C = Conceptual; P = Prototype Architectures and Frameworks, no results are provided; T = Tested Architectures and Frameworks, results are provided. The features annotated with ( \(\ddagger\) ) are associated to Y, P, or N values, that stand for: Y = information are explicitly defined / evaluated; P = information are implicit / stated; N = information are not inferable. Such a categorization of the collected features has been performed according to the DARE criteria, elaborated and proposed by (Kitchenham et al. 2009 ).

4 Review execution

This section details the Perform Review task in Fig.  1 . In particular, it elaborates on the review’s execution, including details on the article collection, selection, and elaboration. The semi-automatic search presented in Sect.  2 resulted in a total of 108 selected articles. The assessment of the primary studies to be finally included in the elaboration phase has been conducted by a total of three reviewers. In particular, the articles have been organized into three equally distributed groups, each of them elaborated by two reviewers (in rotation) with the third one involved in the case of conflict. Footnote 11 Table  2 details the selection assessments, referring to the reviewers with the letters \({\mathcal {A}}\) , \({\mathcal {B}}\) , and \({\mathcal {C}}\) .

The papers have been listed following the collection order and respecting the relevance-based sorting obtained when querying the scientific web collectors. It is possible to notice that the third set of papers recorded a drastic reduction in the acceptance rate. Such a piece of information offers two possible reading keys: (i) the stop criteria has acted too loosely and/or (ii) title & abstract do not mirror the papers’ content properly.

The filtering phase concludes by selecting 38 papers to be elaborated out of the 108 initially collected (21.1% total acceptance rate). In turn, the features presented in section  3.2 have been extracted and collected in a tabular format to facilitate their elaboration and the understanding of possible correlations to be discussed. Nevertheless, in some cases, the extraction of relevant information has been challenging due to the lack of explicit statements (e.g., very few studies have clearly mentioned the limitations of their approaches). To cope with this situation, the reviewers have leveraged their knowledge of the topic to produce a more comprehensive understanding and propose to the reader additional information (rigorously decoupled with the presentation of the results and solely addressed in the discussion).

5 Review results and analysis

In the following, we structure the results of the SLR according to the research questions defined in Sect.  3.1 .

5.1 Demographics

Referring to question SRQ1 , Figs.  2 and 4 show the temporal and geographical distribution of papers targeting agent-based chatbots. Figure  2 reports the primary studies’ distribution over the time-window selected for this study. A slight upward trend can be observed in recent years. Nevertheless, the research field of multi-agent-based chatbots still seems to be a niche area. Indeed, looking at Fig.  4 , the geographical localization of the first authors’ institutions (organized per country) relates to the distribution of research groups in the field of multi-agent systems (i.e., centered in the US and Europe). Finally, Fig.  3 provides a further view on the selected primary studies by grouping the papers per continent.

Total papers per year

Number of papers per continent per year

Number of papers per country

5.2 Application domains

Regarding SRQ2 , we graphically represented in Fig.  5 the selected application domains of the primary studies. It is noticeable that the panorama of the application domains is remarkably broad and diversified. For example, it ranges from education (Alencar and Netto 2014 ) to healthcare (Kökciyan et al. 2021 ) and financing (de Bayser et al. 2018 ). Nevertheless, it appears that personalized assistive purposes have attracted most efforts across domains.

Contributions per application domain

Type of studies

5.3 Intended user classes

Concerning SRQ3 , Fig.  7 shows the distribution of the diverse intended user classes identified by the selected primary studies, which is a direct consequence of the application domains. On the one hand, it is evident that most of the literature operates in the context of education, having either students, tutors, or professors as the main users. On the other hand, although being a minority, a considerable amount of studies is solely conceptual or general (see Fig.  6 ) and does not tackle a specific use case. Overall, the majority ( \(57.89\%\) ) of the primary studies presented some form of prototypes, \(23.69\%\) deal with technical or scientific concepts, and \(18.42\%\) of the selected papers contains extensively tested artifacts.

Number of papers per type of users

5.4 Requirements

Concerning question SRQ4 , we elicited the requirements expressed by the primary studies. We can see the evolution of the main features captured by these requirements in Fig.  9 . We categorized the requirements as follows:

Functional Requirements: requirements affecting the behavior of the platforms (see Table  3 );

Architectural Requirements: requirements stirring the system or the back end of the platforms (see Table  4 );

Front end Requirements: requirements applied to the front end of the platforms (see Table  5 );

Figure  8 depicts the distribution of types of requirements characterizing the primary studies. The authors of the elaborated papers focus primary on functional (41.7%) and architectural (40.0%) requirements. Requirements concerning the front end were only explicitly formalized in 18.3% of the studies.

Type of requirements

Evolution of features in agent-based chatbots according to the requirements

5.5 Objectives of the studies

Investigating SRQ5 , we collected and clustered the objectives of the primary studies as depicted in Fig.  10 . Most of the papers tackle the theoretical foundations of MAS-based chatbots (i.e., nine studies focus primarily on conceptual aspects of the current state of the art or non-concrete systems). Among them, we can mention (Augello et al. 2017 ), where a notion of “social intelligence” for chatbots is defined, and linked to current technologies’ capability to develop social chatbots. Also, (Hung et al. 2009 ) defines a method for an evaluation process to assess the “naturalness” of a chatbot system.

Concerning more practical studies, goal-driven behaviors (e.g., intended to tackle user personalization) have been studied for dietary and entertainment proposes. (Angara et al. 2017 ) describes a chatbot designed to support users in their kitchen by providing recipe recommendations while adhering to their dietary goals, medical conditions, preferences, and available ingredients. Similarly, (Wong et al. 2012 ) describe a goal-oriented virtual chat companion for children with a focus on structured entertainment (e.g., story-telling, collaborative games) and engaging in “free-flowing” dialogue with unstructured responses. Concerning behavioral change, studies such as (Calvaresi et al. 2019 ; Calbimonte et al. 2019 ) target profiling and cravings’ analysis to tailor smoking cessation support, (Calvaresi et al. 2021a ) target the maintenance/improvement of physical balance capabilities with personalized exercises. (Chapman et al. 2019 ), and (Kökciyan et al. 2021 ) demonstrate the development of a chatbot system to help stroke patients manage their care. The system processes data from multiple inputs (e.g., blood pressure monitor, electronic health record) to serve a computational argumentation engine and respond to user queries.

From a different perspective, data-driven behavior has been addressed in contributions including (Agostaro et al. 2005 ; Pilato et al. 2007 ; Augello et al. 2009 ) which deal with the limitations of the conventional, rule-based, data-driven semantics by introducing the paradigm of LSA. Indeed, according to (Landauer et al. 1998 ), LSA allows overcoming rule-based pattern matching limits and introduces an element of intuitiveness by constructing a conceptual space. Another targeted objective is the integration of multiple domain-specific knowledge sources into one chatbot system. For example, (Jiang et al. 2015 ; Augello et al. 2011 ) deal with the integration of different static sources (i.e., vector space model-based indices, XML, relational databases, SPARQL queries, and AIML), while (Pilato et al. 2011 ; Tarau and Figa 2004 ) are intended to manage knowledge dynamically based on the current dialogue context.

While the studies mentioned above are in a user-to-single agent scope, a few studies are in the user to multi-agents (i.e., chatbots) scope. For example, (de Bayser et al. 2017 , 2018 ) address the coordination of multiple bots providing financial advice within the same chat. Their final goal is the moderation of the user-bots’ interaction. Finally, (Calvaresi et al. 2021a ) focused, among other aspects, on the facets of data protection and data privacy.

Primary studies’ objectives

5.6 Technology characterization

Studying SRQ6 , we have classified the primary studies according to the technology readiness level (European Commission 2017 ) (see Table  6 ). In turn, we have analyzed the technologies, architecture, and design principles employed in the primary studies.

Assessing the TRL is a valuable way to measure the maturity of a technology/system. The scale was originally devised by NASA ((Sadin et al. 1989 )) and is nowadays used in many areas in various forms. In this context, we rely on the definition provided by the European Commission in the context of research and innovation projects ((European Commission 2017 )) as shown in Table  6 .

The TRL distribution of the primary studies is depicted in Fig.  11 . It is noticeable that most of the studies are in Levels 3 and 4 (68.1%). This entails that the final outcome of these studies is either a non-validated prototype (TRL 3) or is at the laboratory test stage (TRL 4). Two studies (i.e., (Calvaresi et al. 2019 ) and (Calvaresi et al. 2021a )) are classified as TRL 5. Indeed, such studies have been deployed and validated in real-world health and social-related campaigns.

Technology readiness level distribution of the primary studies

In addition to analyzing the TRL of each study, the front-end and back-end technologies applied in the presented systems were analyzed. All studies with a TRL of 3 and higher were considered. Figure  12 depicts the distribution of the back-end technologies used in the primary studies. The majority (38.7%) of the systems employ Java-based back ends. This prevalence can be related to the wide use of MAS frameworks such as JADE Footnote 12 and MaSMT. Footnote 13 For example, studies such as (Alencar and Netto 2014 ), (de M. Batista et al. 2009 ), and (Bentivoglio et al. 2010 ) rely on JADE and (Hettige and K. 2015 ) implemented the system based on MaSMT. Although not relying on a pre-existing MAS framework, (Pilato et al. 2007 ) and (Tarau and Figa 2004 ) implemented their own ad-hoc Java-based systems. Moreover, (Estes 2011 ) exploit features of the Java Enterprise Edition platform (JavaEE) to develop their chatbot system and (Memon et al. 2018 ) use communication sockets of the Java Standard Edition (Java SE). Several studies use unconventional technologies to develop MAS. For example, (de Bayser et al. 2017 ) use Akka, Footnote 14 an actor-based framework, and (Z. et al. 2016 ) relied on ActiveMQ, Footnote 15 a multi-protocol messaging server.

Overview of utilized back-end technologies

Python-based back ends are 9.7% of the total. In particular, (Jiang et al. 2015 ) and (Calvaresi et al. 2019 ) have developed ad hoc systems, while (Calvaresi et al. 2021a ) rely on the SPADE framework. Footnote 16

Several studies (9.7%) relied on existing proprietary systems. For example, (Kalia et al. 2017 ) and (Angara et al. 2017 )) rely on IBM Watson’s Conversation Platform Footnote 17 and (Zolitschka 2020 ) rely on Aimpulse Spectrum. Footnote 18

A number of studies (9.7%) developed their systems’ back end as ad-hoc solution using JavaScript (i.e., (de Bayser et al. 2018 ), (Thosani et al. 2020 ) and (Bosse. 2021 )).

6.5% of all studies (i.e., (Tarau and Figa 2004 ) and (Bosse. 2021 )) implemented a PROLOG Footnote 19 -based back end. Finally, With a share of 25.8%, a substantial number of studies have developed prototypes but failed to mention details regarding their back end implementation. One such example is (Kökciyan et al. 2021 ). Although the authors specify the human interface, it does not go into detail about how the actual backend is implemented.

Overview of utilized frontend technologies

Figure  13 displays the distribution of the front-end technologies used in the developed chatbot systems. Web-based technologies have received the most attention (31.3%), mostly using JavaScript or JavaServer Pages (JSP) in Java.

Using existing web/mobile messaging platform is a choice undertaken by 15.6% of the studies. In particular, (Calvaresi et al. 2019 ) rely on Facebook Messenger, Footnote 20 (Calvaresi et al. 2021a ) offer Telegram Messenger Footnote 21 among the available interfaces, (Tarau and Figa 2004 ) use Yahoo Instant Messenger (deprecated since 2012), and (Bentivoglio et al. 2010 ) adopt Jabber. Footnote 22

The development of ad hoc solutions accounts for 15.6%. the programming languages involved are Java (e.g., (Hettige and K. 2015 ) or (Tatai et al. 2003 )), C#, and C++ (e.g., (Huang et al. 2008 )).

6.3% of the elaborated solutions’ front ends uses cross-platform frameworks. Such frameworks allow the same code base to be used for web and smartphone app development. For example, The studies used (Thosani et al. 2020 ) use Ionic, Footnote 23 and (Calvaresi et al. 2021a ) offer among the possible interfaces HemerApp, which is written in Flutter. Footnote 24

3.1% of systems used an Android application as front end (e.g., (Kökciyan et al. 2021 )).

Finally, 28.1% of the studies do not mention what technologies are used in their solution or provide only simplistic and non-classifiable descriptions. For example, (de Bayser et al. 2018 ) focuses primarily on the conception of the backend side without mentioning how their human interfacing system was implemented.

5.7 Strengths of the primary studies

Referring to question SRQ7 , the strengths of the primary studies are listed in Table  7 . Among all the strengths, 22% of the strengths are classified as Y, which means that the strengths are explicitly defined and evaluated, 21% are classified as P, indicating that the information is implicitly defined, 57% are classified as N, denoting that the information is not inferrable (see Fig. 14 ). Figure  15 shows, in particular, the classification per strength.

Overview of strength assessment according to the YPN classification. In particular, Y = the information is explicitly defined/evaluated; P = the information is implicit/stated; N = the truthfulness of the information is not inferable

Qualitative assessment of the strengths (Y-P-N criteria). S1 : dynamic update of knowledge base; S2 : adaptability to different domains; S3 : Profiling (according to user behavior); S4 : personalization (according to user input); S5 : reusability of components; S6 : scalability; S7 : performance

5.8 Limitations and solutions of the primary studies

Referring to questions SRQ8 and SRQ9 , the limitations stated in the studies and their proposed solutions were analyzed. Table  8 lists all limitations acknowledged by the authors and their proposed solutions. Only five of the ten papers that point out limitations proposed solutions to address them. As an unfortunate habit, limitations are often overlooked. However, among those who mentioned limitations, it is possible to identify two main categories: architectural and functional. As architectural limitation, we specify limitations that are of technical nature and can be solved by changing the applied architecture or technologies. An example of architectural limitations is (de Bayser et al. 2017 ), which states performance problems when raising the number of participants in a chat group. To solve this problem, they suggest switching to a micro-service architecture. Another example is (Calvaresi et al. 2019 ), emphasizing several limitations of their current system architecture, specifically scaling issues with more complex behaviors, lack of standardized inter-agent communication, and no means of integrating third-party data analysis tools. The solution to these limitations is an entirely new platform based on a MAS. Functional limitations are issues on a functional level that can usually be overcome by exploring alternative approaches to a problem. Examples of functional limitations are (Hettige and K. 2015 ) and (Jiang et al. 2015 ), both of which mention limitations related to semantic processing. The proposed solution of (Hettige and K. 2015 ) is to update the corresponding subsystem, while (Jiang et al. 2015 ) proposes to analyze the user input with domain-independent analyzers (e.g., linguistic analysis or keyboard analysis).

5.9 Future challenges stated in the primary studies

Concerning SRQ10 giving the heterogeneous perspective of the future challenges are rather disparate. However, generally, future challenges can be divided into three categories:

System-related challenges relate to extending already existing functionalities.

Functionality-related challenges refer to new functionality to be implemented.

User-related challenges refer to collecting user experiences (usually in the form of trials).

The studies were analyzed for these three categories. Figure  16 shows the breakdown of the three categories across all studies. With 57.9%, most studies desire to enhance their current system’s stability or expand already implemented functionalities. For example, (Shashaj et al. 2019 ) see improving the system component stability and interoperability with other FIPA Footnote 25 -compliant MAS environments as a future goal, whereas (Calvaresi et al. 2019 ) wish to adapt their architecture to allow distributed computing among several servers to increase performance and to handle agent migration from one server instance to another. A complete list of system-related challenges can be seen in Table  9 . At 28.9%, about one-third of studies are endeavoring to add new functionalities to their existing system. (Vasconcelos et al. 2017 ) attempt to implement more test metrics to test more aspects of a chatbot system, and (Memon et al. 2018 ) seek to expand their chatbot with a graphical user interface and extend its user input capabilities with voice recognition and interpretation. All functionality-related future challenges are listed in Table  10 . 13.2% of all future challenges focus on capturing user feedback. (Alencar and Netto 2014 ) are seeking to test their tutoring system with the help of students and make further improvements to the system based on the feedback collected, and (Kökciyan et al. 2021 ) are conducting two pilot studies with patients to test different aspects of their system. Table  11 lists all user-related challenges stated in the primary studies.

Distribution of future challenge per category

6 Discussion

Analyzing the primary studies emerges that the application of the MAS’ paradigm has slightly increased in the past twenty years, although only moderately. The elaborated works acknowledge the suitability and the intrinsic added value of agent-based systems, including autonomy, goal-setting, and behavior definition. Nevertheless, it appears that these technologies are mostly at an early stage of development. On the one hand, the TRL of most primary studies did not exceed level 3 or 4 (as shown in Fig.  11 ), and it is questionable whether these early-stage systems would be capable of meeting the requirements of a real-world scenario. On the other hand, a few systems have been studied in real-world scenarios (i.e., (Calvaresi et al. 2021a )—testing the developed chatbot in a physical balance-preserving campaign and (Kökciyan et al. 2021 ) – letting both experts and real users analyzing the system. However, it still remains to test such systems in fully operational environments.

Several studies focused on aspects revolving around the management and reconciliation of different knowledge bases. However, only one (Calvaresi et al. 2021a ) has addressed the topic of data privacy and user consent directly. To date, this is a remarkable concern that practitioners have to address imperatively. Indeed, too many studies addressing topics such as user profiling and the processing of user input to enhance chatbot knowledge have either ignored data privacy or not tackled it explicitly. If people are involved, it is of paramount importance to ensure their control over their data. Due to the implementation of more rigid data privacy laws such as GDPR, next-generation systems must have no room to neglect this topic.

The analysis of the technologies’ distribution within the primary studies suggests some trends to be observed. Figure  17 a shows the back-end technologies used over the years. It is possible to notice that Java-based systems have been used extensively. However, since 2015, Python-based systems have emerged, likely due to Python’s prevalence in areas such as machine learning and data science libraries. Moreover, since 2017, the employment of proprietary systems (e.g., IBM Watson) has been increasingly considered. Although initially rather rudimentary, such platforms now offer a wide range of possibilities, such as integrating machine learning modules or extensive analytical capacities. Figure  17 b shows that a shift occurred in the area of front-end technologies too. In addition to the increasing prevalence of web-based solutions, messaging services such as Facebook Messenger or Telegram have become increasingly popular since 2015. Nevertheless, in recent years, the use of cross-platform frameworks became a consistent practice. Cross-platform frameworks such as Ionic or Flutter make it possible to develop front-end solutions for mobile phones and web browsers using a single code base. Moreover, it can be observed a trend to use more complex multi-agent chatbots (e.g., (Bosse. 2021 )) to blend in IoT and micro-services domain with highly scalable multi-agent chatbot networks.

MAS-based chatbot technologies over the years

Most studies have used MAS enabling agents to abstract individual components such as language processing or output composition. (Calvaresi et al. 2019 ) and (Calvaresi et al. 2021a ) have taken a different approach by coupling the users themselves with personalized agents. According to such studies, the goal of this 1:1 relation is to facilitate user profiling, data management, privacy preservation, and personalization. Indeed, by interacting with the user, the respective agent is expected to increase its knowledge and enhance the personalization’s accuracy level over time.

Looking at the evaluation of the strengths of the primary studies in Fig.  14 , it is noticeable that S2 (i.e., adaptability to different domains) and S6 (i.e., scalability) have an above-average number of implicitly defined and evaluated strengths. In the case of S2, this is primarily due to studies having justified their system’s adaptability with the implementation of a single case study to conclude that the system can also be applied to other domains. This is not necessarily a wrong assumption, but the implementation of several distinct scenarios would have been more effective to show this strength explicitly. Compared to S2, S6 is a more generic strength. Since most of the studies are in an early prototype stage, even if the respective systems’ scalability was reported as a strength, this strength was mostly not evaluated. This leads to the question of what methods can be used to evaluate a chatbot platform’s scalability. All studies use the term scalability as a synonym for Size Scalability as defined by (Neuman 1994 ). Size Scalability defines that a system scales easily with the number of users and resources without noticeable loss of performance. To implicitly define this aspect, a load test with several simulated users in which the response times and hardware load of the system are analyzed could be theoretically sufficient.

7 Conclusions

This paper has analyzed the current state of the art of chatbot solutions leveraging the multi-agent approach and agent-based frameworks by performing an SLR. In particular, it employs a well-established methodology characterized by ten structured research questions. Such an investigation focused on aspects including application domains, end-users, requirements, objectives, technology readiness level, designs, strengths, limitations, and future challenges of the solutions found in the literature. Such aspects have been analyzed “per-feature” and overall aggregated in a reconciling discussion. The insights elicited in this work can be beneficial for both theoretical and practical future research.

Program D: 2012 version of ALICE: https://github.com/noelbush-xx/programd.

https://www.apple.com/siri/.

https://developers.google.com/events/io/2012.

Now called Google Assistant: https://assistant.google.com/.

https://www.microsoft.com/en-us/cortana.

http://ieeexplore.ieee.org.

http://www.sciencedirect.com/.

http://dl.acm.org/.

http://citeseerx.ist.psu.edu/index.

http://www.ncbi.nlm.nih.gov/pubmed.

Possible evaluations: Y = Yes, N = No, X = Not sure. If both reviewers agreed on the assessment, no further review was required. However, if a conflict occurred (e.g., Y-X, X-X, X-Y, N-X, X-N), the third reviewer was consulted.

https://jade.tilab.com/ .

https://sourceforge.net/projects/masmt/ .

https://akka.io/ .

https://activemq.apache.org/ .

https://spade-mas.readthedocs.io/en/latest/readme.html .

https://www.ibm.com/cloud/watson-assistant .

https://www.aimpulse.com/ .

https://www.iso.org/standard/21413.html .

https://www.messenger.com/ .

https://telegram.org/ .

https://www.cisco.com/c/en/us/products/unified-communications/jabber/index.html .

https://ionicframework.com/ .

https://flutter.dev/ .

Foundation for Intelligent Physical Agents.

Adamopoulou E, Moussiades L (2020) Chatbots: History, technology, and applications. Machine Learning with Applications 2(100):006

Google Scholar  

Agostaro F, Augello A, Pilato G et al (2005) A conversational agent based on a conceptual interpretation of a data driven semantic space. Congress of the Italian Association for Artificial Intelligence. Springer, New York, pp 381–392

Alencar M, Netto JF (2014) Tutor collaborator using multi-agent system. International Conference on Collaboration Technologies. Springer, New York, pp 153–159

Chapter   Google Scholar  

Angara P, Jiménez M, Agarwal K, et al (2017) Foodie fooderson a conversational agent for the smart kitchen. In: Mindel M, Lyons KA, Wigglesworth J (eds) Proceedings of the 27th Annual International Conference on Computer Science and Software Engineering, CASCON 2017, Markham, Ontario, Canada, November 6-8, 2017. IBM / ACM, pp 247–253, http://dl.acm.org/citation.cfm?id=3172825

Anjomshoae S, Najjar A, Calvaresi D, et al (2019) Explainable agents and robots: Results from a systematic literature review. In: 18th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2019), Montreal, Canada, May 13–17, 2019, International Foundation for Autonomous Agents and Multiagent Systems, pp 1078–1088

Augello A, Pilato G, Vassallo G et al (2009) A semantic layer on semi-structured data sources for intuitive chatbots. 2009 International Conference on Complex. Intelligent and Software Intensive Systems, IEEE, pp 760–765

Augello A, Scriminaci M, Gaglio S, et al (2011) A modular framework for versatile conversational agent building. In: 2011 International Conference on Complex, Intelligent, and Software Intensive Systems, IEEE, pp 577–582

Augello A, Gentile M, Dignum F (2017) An overview of open-source chatbots social skills. International conference on internet science. Springer, New York, pp 236–248

Bentivoglio C, Bonura D, Cannella V et al (2010) Intelligent agents supporting user interactions within self regulated learning processes. J E-learn Knowl Soc 6(2):27–36

Bosse. S (2021) Distributed serverless chat bot networks using mobile agents: A distributed data base model for social networking and data analytics. In: Proceedings of the 13th International Conference on Agents and Artificial Intelligence - Volume 1: ICAART,, INSTICC. SciTePress, pp 398–405, https://doi.org/10.5220/0010319503980405

Calbimonte JP, Calvaresi D, Dubosson F et al (2019) Towards profile and domain modelling in agent-based applications for behavior change. International Conference on Practical Applications of Agents and Multi-Agent Systems. Springer, New York, pp 16–28

Calvaresi D, Dubovitskaya A, Calbimonte JP et al (2018) Multi-agent systems and blockchain: results from a systematic literature review. International conference on practical applications of agents and multi-agent systems. Springer, New York, pp 110–126

Calvaresi D, Calbimonte JP, Dubosson F, et al (2019) Social network chatbots for smoking cessation: agent and multi-agent frameworks. In: 2019 IEEE/WIC/ACM International Conference on Web Intelligence (WI), IEEE, pp 286–292

Calvaresi D, Calbimonte JP, Siboni E et al (2021) Erebots: Privacy-compliant agent-based platform for multi-scenario personalized health-assistant chatbots. Electronics. https://doi.org/10.3390/electronics10060666

Article   Google Scholar  

Calvaresi D, Ibrahim A, Calbimonte JP et al (2021) The evolution of chatbots in tourism: a systematic literature review. Inf Commun Technol Tour 2021:3–16

Chapman M, Balatsoukas P, Ashworth M, et al (2019) Computational argumentation-based clinical decision support. In: Proceedings of the 18th International Conference on Autonomous Agents and MultiAgent Systems. International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC, AAMAS ’19, p 2345-2347

Cui L, Huang S, Wei F, et al (2017) Superagent: A customer service chatbot for e-commerce websites. In: Proceedings of ACL, System Demonstrations, pp 97–102

de M. Batista AF, Marietto MdGB, Barbosa GCdO, et al (2009) Multi-agent systems to build a computational middleware: A chatterbot case study. In: 2009 International Conference for Internet Technology and Secured Transactions,(ICITST), IEEE, pp 1–2

de Bayser MG, Cavalin PR, Souza R, et al (2017) A hybrid architecture for multi-party conversational systems. CoRR arXiv: abs/1705.01214

de Bayser MG, Pinhanez C, Candello H, et al (2018) Ravel: a mas orchestration platform for human-chatbots conversations. In: The 6th International Workshop on Engineering Multi-Agent Systems. Stockholm, Sweden

DeepLink (2022) Deeplink.ai: Artificial intelligence to boost your customer relationship. https://www.deeplink.ai/en/ , Accessed: 2023-02-24

Estes TW (2011) Knowledge discovery agent system and method. US Patent 8,015,143

Etherington D (2014) Amazon echo is a \$199 connected speaker packing an always-on siri-style assistant - techcrunch. https://techcrunch.com/2014/11/06/amazon-echo/ , Accessed: 2023-02-24

European Commission (2017) Appendix g. technology readiness levels (trl). https://ec.europa.eu/research/participants/data/ref/h2020/other/wp/2016_2017/annexes/h2020-wp1617-annex-g-trl_en.pdf

Fadhil A, Gabrielli S (2017) Addressing challenges in promoting healthy lifestyles: the al-chatbot approach. In: Proceedings of the 11th EAI International Conference on Pervasive Computing Technologies for Healthcare, ACM, pp 261–265

Galster M, Weyns D, Tofan D et al (2014) Variability in software systems-a systematic literature review. IEEE Trans Software Eng 40(3):282–306. https://doi.org/10.1109/TSE.2013.56

Guzzoni D (2008) Active: a unified platform for building intelligent applications (phd thesis) pp 1–263. https://doi.org/10.5075/epfl-thesis-3990 , http://infoscience.epfl.ch/record/114758

Hettige B, K. A (2015) Octopus: a multi agent chatbot. In: 8th International Research Conference, KDU, November 2015

Huang HH, Cerekovic A, Tarasenko K et al (2008) Integrating embodied conversational agent components with a generic framework. Multiagent Grid Syst 4(4):371–386

Article   MATH   Google Scholar  

Huddar MG, Sannakki SS, Rajpurohit VS (2021) Attention-based multi-modal sentiment analysis and emotion detection in conversation using RNN. Int J Interact Multim Artif Intell 6(6):112–121. https://doi.org/10.9781/ijimai.2020.07.004

Hung V, Elvir M, Gonzalez A, et al (2009) Towards a method for evaluating naturalness in conversational dialog systems. In: 2009 IEEE international conference on systems, man and cybernetics, IEEE, pp 1236–1241

Jiang R, Banchs RE, Kim S, et al (2015) Configuration of dialogue agent with multiple knowledge sources. In: 2015 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA), IEEE, pp 840–849

Kalia AK, Telang PR, Xiao J, et al (2017) Quark: a methodology to transform people-driven processes to chatbot services. In: International Conference on Service-Oriented Computing, Springer, pp 53–61

Kitchenham B, Pearl Brereton O, Budgen D et al (2009) Systematic literature reviews in software engineering—a systematic literature review. Inf Softw Technol 51(1):7–15. https://doi.org/10.1016/j.infsof.2008.09.009

Kitchenham B, Brereton P, Turner M et al (2010) Refining the systematic literature review process-two participant-observer case studies. Empir Softw Eng 15(6):618–653. https://doi.org/10.1007/s10664-010-9134-8

Kökciyan N, Sassoon I, Sklar E et al (2021) Applying metalevel argumentation frameworks to support medical decision making. IEEE Intell Syst 36(2):64–71. https://doi.org/10.1109/MIS.2021.3051420

Landauer TK, Foltz PW, Laham D (1998) An introduction to latent semantic analysis. Discourse Process 25(2–3):259–284. https://doi.org/10.1080/01638539809545028

Maher ML, Gu N (2002) Design agents in virtual worlds-a user-centred virtual architecture agent. Agents in Design. pp 23–38

Mauldin ML (1994) Chatterbots, tinymuds, and the turing test entering the loebner prize competition. In: Proceedings of the Twelfth AAAI National Conference on Artificial Intelligence. AAAI Press, AAAI’94, p 16-21

Memon Z, Jalbani AH, Shaikh M et al (2018) Multi-agent communication system with chatbots. Mehran Univ Res J Eng Technol 37(3):663

Mori K, Jatowt A, Ishizuka M (2003) Enhancing conversational flexibility in multimodal interactions with embodied lifelike agent. In: Proceedings of the 8th international conference on Intelligent user interfaces, pp 270–272

Mualla Y, Najjar A, Daoud A et al (2019) Agent-based simulation of unmanned aerial vehicles in civilian applications: a systematic literature review and research directions. Futur Gener Comput Syst 100:344–364. https://doi.org/10.1016/j.future.2019.04.051

Neuman B (1994) Scale in distributed systems. Inf. Sci. Inst., Univ. Southern California (ISI/USC), Los Angeles, CA, USA. 68

Noori Z, Bandarl Z, Crockett K (2014) Arabic goal-oriented conversational agent based on pattern matching and knowledge trees. In: Proceedings of the World Congress on Engineering 2014 Vol I. Newswood/International Association of Engineers, July 2 - 4, 2014, London, UK. ISSN 2078-0958, https://e-space.mmu.ac.uk/id/eprint/609597

Palmarini R, Erkoyuncu JA, Roy R et al (2018) A systematic review of augmented reality applications in maintenance. Robot Comput-Integr Manuf 49:215–228

Pereira J, Díaz Ó (2019) Using health chatbots for behavior change: a mapping study. J Med Syst 43(5):135

Pilato G, Augello A, Vassallo G, et al (2007) Sub-symbolic semantic layer in cyc for intuitive chat-bots. In: International Conference on Semantic Computing (ICSC 2007), IEEE, pp 121–128

Pilato G, Augello A, Gaglio S (2011) A modular architecture for adaptive chatbots. In: 2011 IEEE Fifth International Conference on Semantic Computing, IEEE, pp 177–180

Radianti J, Majchrzak TA, Fromm J et al (2020) A systematic review of immersive virtual reality applications for higher education: design elements, lessons learned, and research agenda. Comput Educ 147(103):778

Rao SBP, Agnihotri M, Babu Jayagopi D (2021) Improving asynchronous interview interaction with follow-up question generation. Int J Interact Multim Artif Intell 6:79–89. https://doi.org/10.9781/ijimai.2021.02.010

Rollo C (1997) jabberwacky - about thoughts - an artificial intelligence ai chatbot, chatterbot or chatterbox, learning ai, database, dynamic - models way humans learn - simulate natural human chat - interesting, humorous, entertaining. http://www.jabberwacky.com/j2about , Accessed: 2022-01-10

Sadin SR, Povinelli FP, Rosen R (1989) The nasa technology push towards future space mission systems. Acta Astronaut 20:73–77. https://doi.org/10.1016/0094-5765(89)90054-4

Shashaj A, Mastrorilli F, Stingo M et al (2019) An industrial multi-agent system (mas) platform. International Conference on P2P Parallel, Grid, Cloud and Internet Computing, Springer, New York. pp 221–233

Tarau P, Figa E (2004) Knowledge-based conversational agents and virtual storytelling. In: Proceedings of ACM symposium on Applied computing, pp 39–44

Tatai G, Csordás A, Kiss Á, et al (2003) The chatbot who loved me. In: Proc. ECA Workshop of AAMAS, Melbourne, Australia

Thosani P, Sinkar M, Vaghasiya J, et al (2020) A self learning chat-bot from user interactions and preferences. In: 4th International Conference on Intelligent Computing and Control Systems (ICICCS), IEEE, pp 224–229

Vasconcelos M, Candello H, Pinhanez C, et al (2017) Bottester: testing conversational systems with simulated users. In: Proceedings of the XVI Brazilian Symposium on Human Factors in Computing Systems, pp 1–4

Voigt P, Von dem Bussche A (2017) The eu general data protection regulation (gdpr). A Practical Guide, 1st Ed, vol 10, issue 3152676. Springer International Publishing, Cham. pp 10–5555

Weizenbaum J (1966) Eliza-a computer program for the study of natural language communication between man and machine. Commun ACM 9(1):36–45

Winkler R (2018) Söllner M (2018) Unleashing the potential of chatbots in education: a state-of-the-art analysis. Acad Manag Proc 15:903. https://doi.org/10.5465/AMBPP.2018.15903abstract

Wong W, Cavedon L, Thangarajah J et al (2012) Flexible conversation management using a bdi agent approach. International Conference on Intelligent Virtual Agents. Springer, New York, pp 464–470

Xu A, Liu Z, Guo Y, et al (2017) A new chatbot for customer service on social media. In: Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems, ACM, pp 3506–3510

Yang ECL, Khoo-Lattimore C, Arcodia C (2017) A systematic literature review of risk and gender research in tourism. Tour Manage 58:89–100

Zhao T, Lee K, Eskenazi M (2016) The dialport portal: grouping diverse types of spoken dialog systems. In: Workshop on Chatbots and conversational agents

Zolitschka JF (2020) A novel multi-agent-based chatbot approach to orchestrate conversational assistants. International Conference on Business Information Systems. Springer, New York, pp 103–117

Download references

Acknowledgements

The work is supported by: ◇ HES-SO RCSO ISNet PERSA project: Computational Persuasion in eHealth Support Applications. ◇ Chist-Era grant CHIST-ERA19-XAI-005, and by (i) the Swiss National Science Foundation (G.A. 20CH21\_195530), (ii) the Italian Ministry for Universities and Research, (iii) the Luxembourg National Research Fund (G.A. INTER/CHIST/19/14589586), (iv) the Scientific, and Research Council of Turkey (TÜBİTAK, G.A. 120N680) ◇ The authors thank Diego Collarana from Universidad Privada Boliviana for his insights. This collaboration was supported by the Research Partnership Grant RPG2106 funded by the Swiss Leading House for Latin America.

Open access funding provided by University of Applied Sciences and Arts Western Switzerland (HES-SO).

Author information

Authors and affiliations.

University of Applied Sciences and Arts Western Switzerland, Sierre, Switzerland

Davide Calvaresi, Stefan Eggenschwiler, Michael Schumacher & Jean-Paul Calbimonte

Université de Technologie de Belfort Montbéliard, Sevenan, France

Yazan Mualla

The Sense Innovation and Research Center, Sion, Switzerland

Jean-Paul Calbimonte

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Davide Calvaresi .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Calvaresi, D., Eggenschwiler, S., Mualla, Y. et al. Exploring agent-based chatbots: a systematic literature review. J Ambient Intell Human Comput 14 , 11207–11226 (2023). https://doi.org/10.1007/s12652-023-04626-5

Download citation

Received : 23 March 2022

Accepted : 02 May 2023

Published : 01 June 2023

Issue Date : August 2023

DOI : https://doi.org/10.1007/s12652-023-04626-5

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Agent-based chatbot
• Agents chatbots
• Conversational agents
• Find a journal
• Publish with us
• Track your research

The best AI tools for research papers and academic research (Literature review, grants, PDFs and more)

As our collective understanding and application of artificial intelligence (AI) continues to evolve, so too does the realm of academic research. Some people are scared by it while others are openly embracing the change. 

Make no mistake, AI is here to stay!

Instead of tirelessly scrolling through hundreds of PDFs, a powerful AI tool comes to your rescue, summarizing key information in your research papers. Instead of manually combing through citations and conducting literature reviews, an AI research assistant proficiently handles these tasks.

These aren’t futuristic dreams, but today’s reality. Welcome to the transformative world of AI-powered research tools!

The influence of AI in scientific and academic research is an exciting development, opening the doors to more efficient, comprehensive, and rigorous exploration.

This blog post will dive deeper into these tools, providing a detailed review of how AI is revolutionizing academic research. We’ll look at the tools that can make your literature review process less tedious, your search for relevant papers more precise, and your overall research process more efficient and fruitful.

I know that I wish these were around during my time in academia. It can be quite confronting when trying to work out what ones you should and shouldn’t use. A new one seems to be coming out every day!

Here is everything you need to know about AI for academic research and the ones I have personally trialed on my Youtube channel.

Best ChatGPT interface – Chat with PDFs/websites and more

I get more out of ChatGPT with HeyGPT . It can do things that ChatGPT cannot which makes it really valuable for researchers.

Use your own OpenAI API key ( h e re ). No login required. Access ChatGPT anytime, including peak periods. Faster response time. Unlock advanced functionalities with HeyGPT Ultra for a one-time lifetime subscription

AI literature search and mapping – best AI tools for a literature review – elicit and more

Harnessing AI tools for literature reviews and mapping brings a new level of efficiency and precision to academic research. No longer do you have to spend hours looking in obscure research databases to find what you need!

AI-powered tools like Semantic Scholar and elicit.org use sophisticated search engines to quickly identify relevant papers.

They can mine key information from countless PDFs, drastically reducing research time. You can even search with semantic questions, rather than having to deal with key words etc.

With AI as your research assistant, you can navigate the vast sea of scientific research with ease, uncovering citations and focusing on academic writing. It’s a revolutionary way to take on literature reviews.

• Elicit –  https://elicit.org
• Supersymmetry.ai: https://www.supersymmetry.ai
• Semantic Scholar: https://www.semanticscholar.org
• Connected Papers –  https://www.connectedpapers.com/
• Research rabbit – https://www.researchrabbit.ai/
• Laser AI –  https://laser.ai/
• Litmaps –  https://www.litmaps.com
• Inciteful –  https://inciteful.xyz/
• Scite –  https://scite.ai/
• System –  https://www.system.com

If you like AI tools you may want to check out this article:

• How to get ChatGPT to write an essay [The prompts you need]

AI-powered research tools and AI for academic research

AI research tools, like Concensus, offer immense benefits in scientific research. Here are the general AI-powered tools for academic research. 

These AI-powered tools can efficiently summarize PDFs, extract key information, and perform AI-powered searches, and much more. Some are even working towards adding your own data base of files to ask questions from. 

Tools like scite even analyze citations in depth, while AI models like ChatGPT elicit new perspectives.

The result? The research process, previously a grueling endeavor, becomes significantly streamlined, offering you time for deeper exploration and understanding. Say goodbye to traditional struggles, and hello to your new AI research assistant!

• Bit AI –  https://bit.ai/
• Consensus –  https://consensus.app/
• Exper AI –  https://www.experai.com/
• Hey Science (in development) –  https://www.heyscience.ai/
• Iris AI –  https://iris.ai/
• PapersGPT (currently in development) –  https://jessezhang.org/llmdemo
• Research Buddy –  https://researchbuddy.app/
• Mirror Think – https://mirrorthink.ai

AI for reading peer-reviewed papers easily

Using AI tools like Explain paper and Humata can significantly enhance your engagement with peer-reviewed papers. I always used to skip over the details of the papers because I had reached saturation point with the information coming in. 

These AI-powered research tools provide succinct summaries, saving you from sifting through extensive PDFs – no more boring nights trying to figure out which papers are the most important ones for you to read!

They not only facilitate efficient literature reviews by presenting key information, but also find overlooked insights.

With AI, deciphering complex citations and accelerating research has never been easier.

• Open Read –  https://www.openread.academy
• Chat PDF – https://www.chatpdf.com
• Explain Paper – https://www.explainpaper.com
• Humata – https://www.humata.ai/
• Lateral AI –  https://www.lateral.io/
• Paper Brain –  https://www.paperbrain.study/
• Scholarcy – https://www.scholarcy.com/
• SciSpace Copilot –  https://typeset.io/
• Unriddle – https://www.unriddle.ai/
• Sharly.ai – https://www.sharly.ai/

AI for scientific writing and research papers

In the ever-evolving realm of academic research, AI tools are increasingly taking center stage.

Enter Paper Wizard, Jenny.AI, and Wisio – these groundbreaking platforms are set to revolutionize the way we approach scientific writing.

Together, these AI tools are pioneering a new era of efficient, streamlined scientific writing.

• Paper Wizard –  https://paperwizard.ai/
• Jenny.AI https://jenni.ai/ (20% off with code ANDY20)
• Wisio – https://www.wisio.app

AI academic editing tools

In the realm of scientific writing and editing, artificial intelligence (AI) tools are making a world of difference, offering precision and efficiency like never before. Consider tools such as Paper Pal, Writefull, and Trinka.

Together, these tools usher in a new era of scientific writing, where AI is your dedicated partner in the quest for impeccable composition.

• Paper Pal –  https://paperpal.com/
• Writefull –  https://www.writefull.com/
• Trinka –  https://www.trinka.ai/

AI tools for grant writing

In the challenging realm of science grant writing, two innovative AI tools are making waves: Granted AI and Grantable.

These platforms are game-changers, leveraging the power of artificial intelligence to streamline and enhance the grant application process.

Granted AI, an intelligent tool, uses AI algorithms to simplify the process of finding, applying, and managing grants. Meanwhile, Grantable offers a platform that automates and organizes grant application processes, making it easier than ever to secure funding.

Together, these tools are transforming the way we approach grant writing, using the power of AI to turn a complex, often arduous task into a more manageable, efficient, and successful endeavor.

• Granted AI – https://grantedai.com/
• Grantable – https://grantable.co/

Free AI research tools

There are many different tools online that are emerging for researchers to be able to streamline their research processes. There’s no need for convience to come at a massive cost and break the bank.

The best free ones at time of writing are:

• Elicit – https://elicit.org
• Connected Papers – https://www.connectedpapers.com/
• Litmaps – https://www.litmaps.com ( 10% off Pro subscription using the code “STAPLETON” )
• Consensus – https://consensus.app/

Wrapping up

The integration of artificial intelligence in the world of academic research is nothing short of revolutionary.

With the array of AI tools we’ve explored today – from research and mapping, literature review, peer-reviewed papers reading, scientific writing, to academic editing and grant writing – the landscape of research is significantly transformed.

The advantages that AI-powered research tools bring to the table – efficiency, precision, time saving, and a more streamlined process – cannot be overstated.

These AI research tools aren’t just about convenience; they are transforming the way we conduct and comprehend research.

They liberate researchers from the clutches of tedium and overwhelm, allowing for more space for deep exploration, innovative thinking, and in-depth comprehension.

Whether you’re an experienced academic researcher or a student just starting out, these tools provide indispensable aid in your research journey.

And with a suite of free AI tools also available, there is no reason to not explore and embrace this AI revolution in academic research.

We are on the precipice of a new era of academic research, one where AI and human ingenuity work in tandem for richer, more profound scientific exploration. The future of research is here, and it is smart, efficient, and AI-powered.

Before we get too excited however, let us remember that AI tools are meant to be our assistants, not our masters. As we engage with these advanced technologies, let’s not lose sight of the human intellect, intuition, and imagination that form the heart of all meaningful research. Happy researching!

Thank you to Ivan Aguilar – Ph.D. Student at SFU (Simon Fraser University), for starting this list for me!

Dr Andrew Stapleton has a Masters and PhD in Chemistry from the UK and Australia. He has many years of research experience and has worked as a Postdoctoral Fellow and Associate at a number of Universities. Although having secured funding for his own research, he left academia to help others with his YouTube channel all about the inner workings of academia and how to make it work for you.

Thank you for visiting Academia Insider.

We are here to help you navigate Academia as painlessly as possible. We are supported by our readers and by visiting you are helping us earn a small amount through ads and affiliate revenue - Thank you!

2024 © Academia Insider

Conversational AI: Chatbots

Ieee account.

• Change Username/Password
• Update Address

Purchase Details

• Payment Options
• Order History
• View Purchased Documents

Profile Information

• Communications Preferences
• Profession and Education
• Technical Interests
• US & Canada: +1 800 678 4333
• Worldwide: +1 732 981 0060
• Contact & Support
• About IEEE Xplore
• Accessibility
• Terms of Use
• Nondiscrimination Policy
• Privacy & Opting Out of Cookies

A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity. © Copyright 2024 IEEE - All rights reserved. Use of this web site signifies your agreement to the terms and conditions.

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings

Preview improvements coming to the PMC website in October 2024. Learn More or Try it out now .

• Advanced Search
• Journal List
• Elsevier - PMC COVID-19 Collection

An intelligent Chatbot using deep learning with Bidirectional RNN and attention model

This paper shows the modeling and performance in deep learning computation for an Assistant Conversational Agent (Chatbot). The utilization of Tensorflow software library, particularly Neural Machine Translation (NMT) model. Acquiring knowledge for modeling is one of the most important task and quite difficult to preprocess it. The Bidirectional Recurrent Neural Networks (BRNN) containing attention layers is used, so that input sentence with large number of tokens (or sentences with more than 20–40 words) can be replied with more appropriate conversation. The dataset used in the paper for training of model is used from Reddit. The model is developed to perform English to English translation. The main purpose of this work is to increase the perplexity and learning rate of the model and find Bleu Score for translation in same language. The experiments are conducted using Tensorflow using python 3.6. The perplexity, leaning rate, Bleu score and Average time per 1000 steps are 56.10, 0.0001, 30.16 and 4.5 respectively. One epoch is completed at 23,000 steps. The paper also study MacBook Air as a system for neural network and deep learning.

1. Introduction

The Chatbot can be defined as a software which help humans to make coherent conversation with machine using natural language like English, etc. The conversation can be engaging at times with large vocabularies and broad range of conversational topics. Recently, the usage of deep learning has increased in industry and Chatbot is one of its application [1] , [2] , [3] . Fig. 1 shows user using the Chatbot for its various application. This paper will help to create open-domain Chatbot, which can be later subjected to a particular domain, if needed as shown in Fig. 1 . It can be done by making changes in dataset, which means training model with particular domain knowledge. Due to open domain nature of the Chatbot, it can be used in making Artificial Intelligence Assistant which can make real life conversation with its user in any topic and situation. To make deep learning utilized by everyone, a major deep learning library Tensorflow is implemented by Google [4] and made available for use as an open source. Tensorflow [5] is Python-friendly library bundled with machine learning and deep learning (neural network) models and algorithms. The paper shows the formation of Chatbot by Neural Machine Translation (NMT) model which is improvement on sequence-to-sequence model. The Chatbot use Bidirectional Recurrent Neural Network (BRNN) [6] . The BRNN was chosen, as conversation or input to the Chatbot is dynamic, which means the length of input is unfixed. The BRNN is also supported by attention mechanism [7] , [8] , which help to further increase capacity of model to remember longer sequence of sentences. The concept of Bidirectional Recurrent Neural Network, can be understand by taking two independent Recurrent Neural Network (RNN) [9] together, sending signals through their layer in opposite directions. So BRNN can be seen as neural network connecting two hidden layers in opposite directions to a single output. This helps the network to have both forward and backward information at every step, i.e. to receive information from both past and future states. The input is fed in one direction in normal time order, and the other, in reverse order. The concept of Extended Long Short Term Memory (ELSTM) [10] can also be used, with Dependent BRNN (DBRNN), as it help to increase the result by 30% on labeled data. The training of the BRNN is done in a same way as RNN, as two bidirectional neurons do not interact with one another. When forward and backward passes are done [11] , then only weights are updated.

Chatbot applications across various domains.

The Fig. 2 shows the general BRNN architecture with the hidden states for forward and backward direction. The variable ‘O’, ‘I’ and ‘H’ means ‘Output’, ‘Input’ and ‘Hidden’ states respectively. The values {X1, X2,….,Xn} are input signals to the network and values {Y1, Y2,…. , Yn} are computed output signals from the network.

Basic Bidirectional Recurrent Neural Network Architecture.

1.1. System details

This paper also evaluates MacBook as a system for deep learning. Table 1 shows the system specification and other software details like operating system and version of TensorFlow used. Also the technology is getting upgraded every day, even if we take Central Processing Unit (CPU) and Graphics Processing Unit (GPU), which are becoming faster [12] . The system used is 2017 MacBook Air series by Apple. The laptop was at room temperature all the time of training the model.

System and Software Specification.

2. Literature review

There have been increase in development of conversational agent systems in commercial market like in retail, banking and education sectors. The research is being done, to improve accuracy and make conversation between Chatbot and user as close to real world conversations. Apart from traditional rule based technique, used earlier in Chatbot development and other straightforward machine learning algorithm, advance concepts and techniques like Natural Language Processing (NLP) techniques and Deep Learning Techniques like Deep Neural Network (DNN) and Deep Reinforcement Learning (DRL) are being used to model and train the Chatbot systems. In early days, the translation was done by breaking up source sentences into multiple token or chunks and then translated phrase-by-phrase. Sequence-to-Sequence has been a popular model based on neural network and deep learning, used in Neural Machine Translation [13] . It is used in tasks like machine translation, speech recognition, and text summarization. Sequence-to-Sequence model has an encoder and a decoder, both having Vanilla RNN [14] by default. The encoder take source sentence as input to build a thought vector. A thought vector is a vector space containing sequence of numbers which represent meaning of the sentence. Then finally, decoder process the thought vector fed to it and emit a translation called a target sequence or sentence. But Vanilla RNN fails when long sequence of sentences are fed to model, as information needs to be remembered. This information frequently becomes larger for bigger datasets and create bottleneck for RNN networks. The variation in RNN has to be used like BRNN with Attention mechanism or Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) [15] to handle failure in longer sequences. To understand meaning behind the sentence, the intention, facts and emotions described in the sentence, it must be analyzed. In [3] , the statistical difference was 63.33% to identify two groups or personalities on basis of their sentiments. The technique of NLP and machine learning helps to deeply analyze the sentence sentiments and make comfortable environment for humans to make conversation with machine. If dataset is in text, sentiment features can be classified as text level or document level [16] . The deep learning methods like Convolutional Neural Network (CNN) [17] and RNN are used in document sentiment classification, as it is the difficult one out of the two above mentioned classification. Also developing Chatbot for young adolescent, engaging them in their most preferred channel of communication, that are smartphones and successfully helping them to adult focused care. In [2] , the engagement time was 97%. The Kyoto-NMT is an open-source implementation of NMT paradigm [18] . It used chained, a Deep learning Framework. It has used two layer of LSTM with Attention model in its Recurrent Neural Network. It has also used whitespaces for making token in data preparation. The vector size used by it is of 1000. The training data is a sentence-aligned parallel corpus that is in utf-8 text files: one for source language sentences and the other for target language sentences. In this, a JSON file is created, containing all the parameters. SQLite is used for database creation and for keeping track of training progress. The bleu score is computed by applying greedy search on validation dataset. As the real translation was there (Japanese-to-English), the bleu score they reached is of 26.22. The Snowbot is a Chatbot, developed on Tensorflow and MXNet [19] . The Chatbot uses Cornell movie Dialog Corpus with 221,282 QA, size 22 MB and twitter chat with 377,265 QA having emoji in sentences, size 51 MB. The Chatbot was developed in desktop having NVIDIA GTX 1080 8 GB and 5 GB RAM. The neural network contains two layer of LSTM with 256 hidden units. The vocabulary size was of 50000. The model completed 7 epochs in 1 h. The perplexity was 90 and 135 for Cornell and Twitter dataset respectively.

2.1. Chatbot performance metrics across industries

Currently, there are many performance metrics, and certain measurement standards are followed across industry for Chatbot [20] . Different organizations need Chatbot according to the nature of their work and market surrounding it. One of the most important performance metric for Chatbot is the structure and the length of its conversation. The length of output sentence must be appropriate and in context to the conversation being done. Shorter and simple the structure of sentence in output, faster the solution, does increase the customer satisfaction rate. To understand this metric with an example of banking sector, in banks, the Chatbot is mainly used to guide the user through the bank’s policy, schemes and other customer inquiries about their account. This serve user to perform their tasks quicker, and also lower the human call assistance, thus cutting cost in the service. The consumer satisfaction also bring second metric-the retention rate, very important for Chabot success. The companies aim for significantly high retention rate, indicating customer satisfaction. The automated calls and Chatbot messengers are being used to replace other communication mediums (i.e. lowering call volumes by humans). The retention rate increases when Chatbot are more trained to support the user in managing their account without speaking to a human assistant. Another metric is the ability of the Chatbot, to produce the personalized reply to the user. This means that the Chatbot should take in the source sentence, understand and analyze it, and produce an output statement mapped to the particular problem or query of the user. The companies try to personalize and customize the output statement according to each user’s need, like bank suggesting user a relevant offer or credit card scheme according to the balance in the user’s account, salary, current loan and its spending history. For example, Erica, which is a Chatbot, is Bank of America’s AI virtual assistant, which combine predictive analysis and NLP, to help its users to access their balance information, tips on better saving the money depending upon their spending habits, making transaction between accounts and schedule meetings at financial centers and banks. The e-market and Retail Chatbots make engaging environment for users to shop. Through their environment, the Chabot transform itself in a personal assistant for assisting in shopping. Unlike banking and financial sector Chatbots, Retail conversational agents are designed to look for higher number of conversational steps by holding the users attention, providing details and encouraging them to browse more and ultimately purchase the product. For instance, Ebay’s ShopBot, help users to find best deals from its list of billion products. It is easy to-talk-to, like a friend, either if one is searching for a specific product or browsing to find something new. The above discussed studies shows network designed for small sized datasets and for short input sentences which are not fit for real life conversation as human tends to speak in longer sentences. To counter real world conversation, model like BRNN is important to know conversation context and references, from past as well as future. Attention mechanism is important attachment to the network as it help to weigh the particular references from the input sentences. Also for evaluating Chatbot performances there are no hard metrics [21] , but parameters like Perplexity, Learning rate and Bleu score can represent as to how close one can approach at the time of training the model.

3. Methodology

The model, BRNN with Attention model not only help in short but also in longer tokens. Attention model help us to remember longer sequences at a time and also help in context problem where both historical and future information is required. In real world as language may not necessarily be in perfect sequence, sometimes one has to use context, hear full conversation before going back and responding to words leading up to that point. Human tends to speak in longer sentences to understand the meaning. This is the reason that makes combination of BRNN and Attention model, perfectly right choice for Chatbots. The BRNN structure forms the acyclic graph as can be seen in Fig. 3 . The signals are of two types: forward and backward. The a → is forward recurrent component and a ← is backward recurrent component. There are input values  X  = { X 1 , X 1 , ⋯ . X n } at timestamp t = { 1 , 2 , ⋯ . n } and predicted values are Y ^ = { Y ^ 1 , Y ^ 2 , ⋯ . Y ^ n }. To make prediction Y ^ a t t i m e t , which is an activation function g(n) , computed as:

where W y is the weight according to the input and magnitude set in the network, a → t & a ← t , are the forward activation at time t and backward activation at time t respectively. The P y is the computed value (or predicted value) from the previous neuron in the direction, information is advancing. The activation function in neural network are the function attached to the node (or neuron), which get triggered when the input value in the current node is relevant in making prediction. There are many types of activation functions, but with BRNN, sigmoid and logistic activation functions are mostly used. For example, in the network shown in Fig. 3 , an input of couple of statements has been given. Statement 1- He said, “Indira went to the market.” and statement 2- He said, “Indira Gandhi was a great Prime Minister.” At an instance, when statement 2 was inserted into the network, at time step 3, where word “Indira” was inserted, Y 3 ^ cell prediction or output signal is checked, and due to bidirectional in nature, the forward information flows from cell 1st and 2nd to 3rd cell, as well as, backward flow of information from 9th cell (through all cells in between) to 3rd, help the cell to predict that ‘Indira’ in statement 2 is Prime Minister and not ‘Indira’ who went to the market, in statement 1. If simple RNN have been used, the output must be according to statement 1, as the network didn’t have the future information.

Source input demonstration into BRNN with hidden layer to describe forward and backward flow of information.

This is because the RNN are unidirectional, i.e. with positive direction of time [22] . In attention mechanism, attention vector is generated by calculating score and then calculated vector is retained in memory, so as to choose between best candidate vectors. The score is calculated by comparing each hidden target with source hidden state. For applying attention mechanism, single directional RNN can be used above BRNN structure. Each cell in the Attention network is given context, as an input. This type of network is also called context-aware attention network [23] . The predicted value from each BRNN cell is taken, combined with value from the previous state (or neuron) of the attention network for calculating the attention value. One can also say that context is weight of the features from different timestamp. The Fig. 4 shows how the weighted value is taken in attention cell, as described above. The context C for cell 1in attention network can be computed as:

where a < t ' > = a → t , a ← t and Y < 1 , t ' > is the value from activation function applied on BRNN for prediction on its each cell, that will be used as weight for context computation. Y < t , t ' > is basically the amount of attention, Y < t > should pay to a < t ' > .

Context-Aware Attention Model architecture.

The S n represents the states of the attention model, where n ∈ W . The S 0 is the primary weight, whose value is set according to the network. The Z ^ t at time step t is the attention value after computation from attention neuron where t ∈ N . The Fig. 4 is the extended structure, which when combined with BRNN in Fig. 3 to produce complete architecture for the deep learning network. This network has quadratic cost for time. As generally, in Chatbot, not much longer sentences or paragraph are used to converse, so cost may be acceptable. Though there are other research in this field to reduce quadratic time cost. The attention mechanism is one of the important methods in deep learning techniques, especially in area of document classification [24] , speech recognition [25] , and in image processing [26] , [27] , [28] .

3.1. Implementation

The procedure for implementing methodology is depicted in Fig. 5 .

Procedure for implementing methodology.

3.1.1. Datasets

The Reddit dataset [29] has been used to make database for the Chatbot. The dataset contain comments of January, year 2015. The format of data is in JSON format. The content of dataset is parent_id, comment_body, score, subreddit, etc. The score is most useful to set the acceptable data criteria as this show that this particular comment is most accurate reply to the parent_comment or parent_body. Subreddit can be used to make some specific type of Chatbot like scientific or other particular domain Bots. A subreddit is a specific online community, and the posts associated with it are dedicated to a particular topic that people write about. The database formed after pre-processing the dataset have size of 2.42 GB. The database contain 10,935,217 rows (i.e. number of parent comment-reply comment pairs).

3.1.2. Preprocessing

Now first, for training the model, database is required. So dataset is converted into a database with fields like parent_id, parent, comment, Subreddit, score and UNIX (to track time). To make data more admissible, take data (comment) which have less than 50 words but more than 1 word (in case reply to parentis empty). Also remove all newline character, ‘[deleted]’ and ‘[removed]’ comments, etc. If data (comment body) is valid according to acceptable criteria and has more score than previously paired comment to parent comment of same parent_id, then replace it. Also if encountered with a comment with no parent comment then it means, it can itself be parent comment to some other comment (i.e., it is main thread comment in Reddit). For database creation, the data have been paired in parent and child comments. Each comment is either a main parent comment or reply comment, but each have parent_id. Each parent comment and its reply comment has same parent_id. The pairs are made in accordance to the parent_id. In creation of database, the parent comment is mapped with its best child or reply comment. Any comment, either a parent or a child, have an acceptance score of two. When encountered with a new comment, if it matches the parent_id of previous entered reply comment to a parent body, then compare it with entered reply comment score. If current comment has better score than existing mapped reply comment, the replacement is done between new and previous reply comment and other associated data. If not the case, then the row remains unchanged. Further, if comment encountered has a parent body which is not yet paired with any reply comment, map the comment with its parent body, else if comment has no parent body, then create a new row for the comment, as the new comment encountered can be a parent to some other reply comment. On creation of database, 10,935,217 parent-reply comment pairs (rows) are created.

3.1.3. Training model

For training after creation of database, rows have to be divided into training data and test data. For both, two files are created (i.e., Parent comment and Reply comment). Training data contains 3,027,254 pairs and Test data contains 5100 pairs. There are also list of protected phrases (e.g.www.xyz.com should be a single token) and blacklisted words, to avoid feeding it to learning network. The training files are fed to multiprocessing tokenizer, as they are CPU intensive. The sentences will be divided into tokens on basis of space and punctuation. Each token will act as vocabulary. For each step, vocabulary size is 15000. The size is appropriate for systems having virtual memory of 4 Gigabytes. The RegX module is used for formulating search pattern for vocabulary. It is faster than standard library and it is basically used to check whether a string contain a specific search pattern. The neural network must be designed as mentioned above. Once the training starts, the main concerned hyperparameters (HParams) in metrics are bleu score (bleu), perplexity (ppl) and learning rate (Lr). Bleu score tells, how good the model is translating a sentence from one language to another language. It should be as high as possible. Perplexity is a measure of the probability distribution, or it tells about model prediction error. Learning rate reflects the model’s learning progress in the network. As in this paper, language at both ends of the model is English, so Perplexity is more useful than bleu score. Learning rate is useful but only when model is trained with large data and for longer period of time. If model is trained for limited period of time or with less data, no significant change in learning rate will be observed.

3.1.4. Result and analysis

Initially, the perplexity before training the model was 16322.15, learning rate was 0.001 and bleu score was 0.00. The average time used by above described system, per 1000 steps is between 4 and 4.5 h. If upper bound of time is taken, then for training machine till 23,000 steps, the system took 103.5 h. The perplexity, learning rate and bleu rate at step 23,000 is 56.10, 0.0001 and 21.67. The maximum value of bleu score model reached was of 30.16 at 18000 th step. The model also passed one epoch at 23000th step. The learning rate is low and negligible, as changes were made externally to the weights in the neural network, once the training started. The performance evaluation is shown in Table 2 .

Performance evaluation.

Also one can draw comparison of the performances of the Chatbot in [18] and in [19] with our result, as reflected in Table 3 .

Performance comparison.

The graph of perplexity and bleu score is shown in Fig. 6 . The other observations in Tensorboard like train_loss, decreases when model starts training, and once if train_loss starts increasing after reaching a minimum point, one should stop training the model, as very less or no change in model performance will occur. Thus, it describe that more and excessive training of model can lead to data loss. The smoothing of all graphs is done at value of 0.96 for better interpretation.

The Perplexity and Bleu Score graph.

Speed graph in Fig. 7 demonstrate the system speed per 1000 steps. As mentioned above, no real translation is going on this Chatbot, but still value should initially increase. There will be decrease in value too at some points as no real translation is taking place. The perplexity should fall in every case. If it doesn’t fall, it means the model is not getting trained properly. Also there will be not significant change in reply of NMT Chatbot. There is speed variation in value of speed of system as the speed depends upon overall task getting performed and other opened-up and running applications. From analysis and experience on system while working on experiment, the MacBook Air is just enough for basic deep learning model training, but not adequate. If one wants to go higher, and train some intermediate and advance model, MacBook Air (2017) hardware is not enough. There are insignificant change in latest model of MacBook Air, for instance MacBook Air (2020) or other MacBook Air series.

The System speed graph (x-axis denotes number of steps and Y-axis denotes system time (in sec).

With help of test file, previously created for validating Chatbot’s reply, Table 4 gives the comparison between the source dataset’s test reply comment and the Chatbot’s (NMT-Chatbot) reply after 23,000 steps of training. The test reply comment means the real world human reply to the test parent comment on Reddit. The NMT-reply is the output from the Chatbot. The eight sentences are randomly picked from the parent comment field of the database created for training. The sentences below have length range between 8 and 30, in terms of words. No punctuation in the sentences shown in Table 4 are added or removed.

Conversation input–output response analysis of referenced user versus NMT-Chatbot reply.

4. Conclusion and future works

A Chatbot using deep learning NMT model with Tensorflow has been developed. The Chatbot architecture was build-up of BRNN and attention mechanism. The Chatbot Knowledge base is open domain, using Reddit dataset and it’s giving some genuine reply. In future, the model will be rewarded on relevant and sentiment appropriate reply. This will involve Deep Reinforcement Learning (DRL) technique. Also the methodology used in implementing and training the chatbot, can be used to train the specific domain chatbot, like scientific, healthcare, security, banking, e-market and educational domain. This approach will help building the chatbot in any domain easier and can improve the existing chatbot based on simple RNN architecture or other neural network by using attention mechanism as above. To implement domain specific chatbot (like healthcare, education, etc.), one can download specific Subreddit, of the particular domain. The future work will also include to build a healthcare Chatbot, guiding patient of diseases like COVID-19 (pandemic), Diabetes, High Blood Pressure and heart, etc. by providing information about the inquired disease, food one can eat and ways to deal with several emergency situations. This Chatbot will be powered by a recommender system too. In this paper, the novel idea was to analyze MacBook Air as a system to study and train deep neural network model. We find the MacBook air as mediocre and basic level system for deep learning. This result can help basic level students or other professionals to choose system wisely before starting with deep learning.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

'ZDNET Recommends': What exactly does it mean?

ZDNET's recommendations are based on many hours of testing, research, and comparison shopping. We gather data from the best available sources, including vendor and retailer listings as well as other relevant and independent reviews sites. And we pore over customer reviews to find out what matters to real people who already own and use the products and services we’re assessing.

When you click through from our site to a retailer and buy a product or service, we may earn affiliate commissions. This helps support our work, but does not affect what we cover or how, and it does not affect the price you pay. Neither ZDNET nor the author are compensated for these independent reviews. Indeed, we follow strict guidelines that ensure our editorial content is never influenced by advertisers.

ZDNET's editorial team writes on behalf of you, our reader. Our goal is to deliver the most accurate information and the most knowledgeable advice possible in order to help you make smarter buying decisions on tech gear and a wide array of products and services. Our editors thoroughly review and fact-check every article to ensure that our content meets the highest standards. If we have made an error or published misleading information, we will correct or clarify the article. If you see inaccuracies in our content, please report the mistake via this form .

How to use ChatGPT to summarize a book, article, or research paper

AI chatbots like ChatGPT can be used to make summarizing long articles, research papers, and books an easier job. If you're tasked with writing a summary for school or work about a body of written text, and you're pinched for time, ChatGPT can help you understand the necessary components.

You should remember that ChatGPT is a tool that can help you further understand a topic, and it may not be in your best interest to have it write your work for you.

Also:   How to make ChatGPT provide sources and citations

If you're a student writing a research paper, someone who is keen to discover more about a lengthy article, or someone who wants to dive into a complicated subject, you can use ChatGPT to simplify the process.

How ChatGPT can create summaries for you

Materials needed : You'll need a device that can connect to the internet, an OpenAI account , and a basic understanding of the article, research paper, or book you want to summarize. 

Also:  This AI chatbot can sum up any PDF and answer any question you have about it

The process should take about one to three minutes.

1. Find your article, paper, or book to summarize

If you need ChatGPT to help summarize an article or research paper, find the body of text online and keep it open in a separate tab.

2. Open your web browser and log in to OpenAI

Open your web browser of choice and type in chat.openai.com/chat . 

Also: How to use ChatGPT to write Excel formulas

Log in or sign up for an account. 

3. Type your request

In the chat box, type in TLDR: followed by the title of the book you want summarized. TLDR stands for too long, didn't read. 

Also: How to use ChatGPT: Everything you need to know

You can also ask ChatGPT: "Summarize [book title]." 

You cannot paste a URL into ChatGPT. If you do so, you'll get a response explaining that the chatbot cannot access websites. If you want ChatGPT to summarize research, you'll need a basic understanding of the topic to ensure you aren't including inaccurate information in your report.

For example, if you're writing a paper about how twin sibling dynamics affect their marriages, you can ask ChatGPT: "Explain how twin relationships affect their marriages."

Also:   The best AI chatbots: ChatGPT and other noteworthy alternatives 

ChatGPT will offer you a few bullet points of both positive and negative effects a twin bond can have on the twins' marriages. You'll still need to read key points of the research, as ChatGPT won't provide you with specific statistics or scientific conclusions. But ChatGPT can help you understand the context surrounding the research you're interested in.

If you're having trouble comprehending specific passages in an article, book, or research paper, you can copy parts of the text and paste them into ChatGPT.

Think of this method as reshuffling the words you're currently reading to help you make more sense of the text in front of you.

This is an accurate summary of 'The Hunger Games: Mockingjay'.

What are ChatGPT's limitations?

If you're using ChatGPT to summarize an article, book, or piece of research, keep in mind that ChatGPT isn't aware of events that occurred after September 2021. 

Also:   4 things Claude AI can do that ChatGPT can't

ChatGPT is a large language model that uses queues and millions of data points to mimic human responses. This form of mimicry is why ChatGPT will answer questions even when it doesn't output the correct answer. So, make sure you're not using any information from ChatGPT without fact-checking it.

If you try to get around this obstacle and provide ChatGPT with an article that contains information post-2021, it might hallucinate. Here, I asked the chatbot to summarize an article about a new app I wrote about , and it made up a few details. 

Lemon8 is a new app from TikTok's parent company, ByteDance. Although the TikTok trend may exist, that's not what the article is about.

Can ChatGPT summarize a PDF?

Copying and pasting a URL into ChatGPT won't yield you the best results. If there is a specific paragraph or sentence you're struggling with, you can copy and paste it into ChatGPT and ask it to explain it to you. Still, it's best to read the PDF and use the chatbot as a summary tool and not as an educator. 

Also:  How to use ChatGPT to write code

If you're looking for an AI chatbot that you can regularly rely on to give you an accurate summary of a PDF, consider using ChatPDF . You can summarize up to three PDFs of up to 120 pages per day, and an upgraded plan is available for $5 per month.

Can ChatGPT summarize an email thread?

Sort of. If you want to copy and paste every single email, ChatGPT can summarize the thread's contents for you. It would be more helpful to scan an email thread yourself and ask ChatGPT to help you write a response based on the key points you know about the conversation. 

Editor's note: We've added additional context to the step concerning ChatGPT summarizing articles by URL. 

More on AI tools

How to use chatgpt, what is chatgpt and why does it matter here's what you need to know, how to use copilot pro to compose, reply to, and summarize emails in outlook.

To revisit this article, visit My Profile, then View saved stories .

• Backchannel
• Newsletters
• WIRED Insider
• WIRED Consulting

Amanda Hoover

Students Are Likely Writing Millions of Papers With AI

Students have submitted more than 22 million papers that may have used generative AI in the past year, new data released by plagiarism detection company Turnitin shows.

A year ago, Turnitin rolled out an AI writing detection tool that was trained on its trove of papers written by students as well as other AI-generated texts. Since then, more than 200 million papers have been reviewed by the detector, predominantly written by high school and college students. Turnitin found that 11 percent may contain AI-written language in 20 percent of its content, with 3 percent of the total papers reviewed getting flagged for having 80 percent or more AI writing. (Turnitin is owned by Advance, which also owns Condé Nast, publisher of WIRED.) Turnitin says its detector has a false positive rate of less than 1 percent when analyzing full documents.

ChatGPT’s launch was met with knee-jerk fears that the English class essay would die . The chatbot can synthesize information and distill it near-instantly—but that doesn’t mean it always gets it right. Generative AI has been known to hallucinate , creating its own facts and citing academic references that don’t actually exist. Generative AI chatbots have also been caught spitting out biased text on gender and race . Despite those flaws, students have used chatbots for research, organizing ideas, and as a ghostwriter . Traces of chatbots have even been found in peer-reviewed, published academic writing .

Teachers understandably want to hold students accountable for using generative AI without permission or disclosure. But that requires a reliable way to prove AI was used in a given assignment. Instructors have tried at times to find their own solutions to detecting AI in writing, using messy, untested methods to enforce rules , and distressing students. Further complicating the issue, some teachers are even using generative AI in their grading processes.

Detecting the use of gen AI is tricky. It’s not as easy as flagging plagiarism, because generated text is still original text. Plus, there’s nuance to how students use gen AI; some may ask chatbots to write their papers for them in large chunks or in full, while others may use the tools as an aid or a brainstorm partner.

Students also aren't tempted by only ChatGPT and similar large language models. So-called word spinners are another type of AI software that rewrites text, and may make it less obvious to a teacher that work was plagiarized or generated by AI. Turnitin’s AI detector has also been updated to detect word spinners, says Annie Chechitelli, the company’s chief product officer. It can also flag work that was rewritten by services like spell checker Grammarly, which now has its own generative AI tool . As familiar software increasingly adds generative AI components, what students can and can’t use becomes more muddled.

Detection tools themselves have a risk of bias. English language learners may be more likely to set them off; a 2023 study found a 61.3 percent false positive rate when evaluating Test of English as a Foreign Language (TOEFL) exams with seven different AI detectors. The study did not examine Turnitin’s version. The company says it has trained its detector on writing from English language learners as well as native English speakers. A study published in October found that Turnitin was among the most accurate of 16 AI language detectors in a test that had the tool examine undergraduate papers and AI-generated papers.

Lauren Goode

William Turton

Eric Ravenscraft

Schools that use Turnitin had access to the AI detection software for a free pilot period, which ended at the start of this year. Chechitelli says a majority of the service’s clients have opted to purchase the AI detection. But the risks of false positives and bias against English learners have led some universities to ditch the tools for now. Montclair State University in New Jersey announced in November that it would pause use of Turnitin’s AI detector. Vanderbilt University and Northwestern University did the same last summer.

“This is hard. I understand why people want a tool,” says Emily Isaacs, executive director of the Office of Faculty Excellence at Montclair State. But Isaacs says the university is concerned about potentially biased results from AI detectors, as well as the fact that the tools can’t provide confirmation the way they can with plagiarism. Plus, Montclair State doesn’t want to put a blanket ban on AI, which will have some place in academia. With time and more trust in the tools, the policies could change. “It’s not a forever decision, it’s a now decision,” Isaacs says.

Chechitelli says the Turnitin tool shouldn’t be the only consideration in passing or failing a student. Instead, it’s a chance for teachers to start conversations with students that touch on all of the nuance in using generative AI. “People don’t really know where that line should be,” she says.

You Might Also Like …

In your inbox: The best and weirdest stories from WIRED’s archive

Jeffrey Epstein’s island visitors exposed by data broker

8 Google employees invented modern AI. Here’s the inside story

The crypto fraud kingpin who almost got away

It's shadow time! How to view the solar eclipse, online and in person

Will Knight

Kate Knibbs

Matt Burgess

Reece Rogers

Estelle Erasmus

20+ ChatGPT Alternatives You Can Try For Free (2024)

Discover the top ChatGPT alternatives for writing, translation, coding, and research to meet your specific needs in our curated list.

While ChatGPT is a popular and powerful conversational AI chatbot, there are other options with better service offerings, improved AI models,  unique features/capabilities, and suited to a specific need.

Which ChatGPT Alternative Should You Use?

Choosing the perfect ChatGPT alternative for your needs out of the sea of options can be overwhelming.

To help you out, we placed the AI chatbots can into four main categories, depending on their distinct features and what you want to achieve.

Here's a breakdown of the four  ChatGPT alternative categories and the options you can try in each category:

• If you need an LLM-specific AI chatbot built on a custom AI model different from OpenAI's GPT,  try Claude , Gemini , Mistral , or Llama .
• Go with Monica or Quora's Poe for a general-purpose, all-in-one AI assistant powered by multiple AI models and featuring an AI Bots Hub (marketplace) where you can use different personalized bots for your needs with one account.
• There are Special AI Apps built by big tech companies that can replace ChatGPT for certain functions such as Microsft's Copilot , Github's Copilot , and Amazon Codewhisperer , and Grok (by Elon Musk's xAI) .
• Whether it's for academic research, teaching, or copywriting, there are plenty of AI chatbot options suited to your specific needs in the Special Tasks category. They include Perplexity AI, WordAI,  Elicit (for academic research), Jasper (for marketing content), Learnt.ai (for educational tasks), and AnonChatGPT (for using ChatGPT anonymously).

The Best ChatGPT Alternatives You Can Try

Here’s our list of the best ChatGPT alternatives you can get started with:

Monica is our all-in-one AI assistant built to help you generate text, chat, search, translate, and simplify your tasks. Think of it as ChatGPT on steroids.

There are many reasons why we believe Monica is the best ChatGPT alternative on the market.

With our AI assistant in your toolset, you get an all-in-one integrated platform that can seamlessly transition between different large language models (LLMs).

So depending on the task you want to complete, you can choose between well-known AI models including GPT-4 , Claude 3 Opus , DALL-E 3, LLAMA, Gemini , and Mistral .

Monica offers cross-platform convenience and usability, through our mobile apps, desktop apps, and browser extensions (for Google Chrome and Edge). This allows you to easily access the powerful suite of chatbot features while browsing the web, at any time, and on any device (Mac, Windows, iOS, and Android).

OpenAI’s GPT model tends to hallucinate sometimes and this is one big concern using ChatGPT. Monica aims to solve this by referencing sources from where an answer was pulled via hyperlinked footnotes so you can validate whatever information it generates.

Finally, Monica enables a smarter, personalized AI chat experience by having its own Prompt and Bots Marketplace. With this development, you can get extensive customization options, tailored interaction experiences, custom-made prompts, and personalized services.

Need an all-in-one AI assistant that can seamlessly write code, craft essays, summarize content, search the internet intelligently, and even generate AI images ?

Monica is a must-have, recommended ChatGPT alternative.

• Creator : Monica
• Pricing : Free, paid plans start from $9.99/month

Gemini (formerly Bard)

Gemini is Google's answer to ChatGPT.

Formerly known as Bard, the AI chatbot connects with Google's extensive search index to offer up-to-date information when prompted.

Gemini has an intuitive interface that’s easy to use. Like ChatGPT, you’ll enter a text prompt to generate relevant answers, search information from the web, or even generate code.

It’s a recommended alternative to ChatGPT, especially if you want to look up real-time information on the web.

• Creator : Google
• Pricing : Free

Copilot by Microsoft

Copilot is Microsoft’s AI chatbot that runs on a combination of the GPT-4 AI model and the Bing search engine index.

Formerly referred to as Bing AI, the chatbot provides real-time answers to prompts by accessing its AI database and the internet.

To tweak how the bot responds, you can choose from three conversation styles: Balanced , Precise , or Creative .

For each response generated, you’ll find source links below that you can click through to get additional information. And at the click of a button, you can easily switch from Copilot to the normal Bing search engine.

• Creator: Microsoft

Try Copilot

Perplexity AI

Perplexity is another ChatGPT alternative that generates short responses and content in a conversational style.

With a simple ChatGPT-like user interface, the chatbot generates information to prompt by leveraging the OpenAI model. It also pulls in information from popular sites like Wikipedia, Amazon, and LinkedIn.

• Creator: Perplexity
• Pricing: Free. Paid plans start from $20 per month

Try Perplexity

Claude is a next-generation AI assistant built by Anthropic to help you generate answers, search for information, and do your best work.

This ChatGPT alternative was created mainly for work and has been trained to be safe, accurate, and secure.

Claude 3 is the latest iteration of Anthropic’s AI model family and this new model sets new industry benchmarks across various cognitive tasks. Claude 3 has three separate AI models: Haiku, Sonnet, and Opus, with each offering distinct performance capabilities.

• Creator: Anthropic
• Pricing: Free. Claude Pro costs $20 per month

Jasper Chat

Jasper.ai is a conversational AI tool that provides a suite of AI writing tools perfect for marketers and content creators to escape writer's block.

Like ChatGPT, the AI writing tool assists marketing teams with writing inspiration by generating articles, ad copy, social media content, video scripts, and even AI images. It could also be used for sales, customer service, and marketing-related tasks.

To generate content, Jasper.ai leverages popular AI models: GPT-4 and Claude.

• Creator: Jasper
• Pricing: Free trial. Paid plans start from $49 per month

Try Jasper Chat

Chatsonic is a ChatGPT alternative designed to provide conversational responses along with real-time data, image, and voice searches. It was built to solve the limitations of ChatGPT at a 20% cost savings.

The conversational AI chatbot is powered by GPT-4 and is integrated with Google Knowledge Graph to generate factual and relevant content on any given real-time topic.

Plus, Chatsonic has a text-to-image AI art generator, Chrome extension, and persona mode (with 13+ different potentialities) to enrich the chat experience.

• Creator: Writesonic
• Pricing: Free trial available. Paid plans start from $16.67 per month

Try Chatsonic

Copilot by GitHub

Copilot is the world’s most widely adopted AI developer tool and a recent addition to GitHub.

The AI tool was built on OpenAI Codex's GPT-3 model, to boost developer productivity and accelerate software development.

Copilot offers multi-language coding support and lets you generate syntax in up to 12 programming languages (including JavaScript, Go, PHP, Ruby, and more).

One good thing is it has compatibility with various popular development environments such as JetBrains, VS Code, Neovim, and more.

• Creator: Github
• Pricing: Free and Pro plan for $19 per month

Try CoPilot by Github

Character.AI

Character AI allows you to enjoy rich conversational experiences through AI personas/characters.

With the AI Chatbot, you can choose to act as and converse with various real-life and fictional personalities. Some of these characters on the platform include Sam Altman, Elon Musk, Bill Gates, Harry Potter, and more.

For instance, you can ask Elon Musk's AI character why he bought Twitter and the tool will return an appropriate response. You can also build your character and make it public.

This ChatGPT alternative was built on neural language models and is fun to use if you love conversational AI chatbots.

• Creator: Character Technologies
• Pricing: Free but sign-up required

Try Character AI

AnonChatGPT

This is simply Incognito ChatGPT . If you ever want to use OpenAI’s chatbot without creating an account or logging in, AnonChatGPT was made for you.

Built on OpenAI's GPT-3, the chatbot routes all your prompts to OpenAI's servers and returns the necessary responses. This way, your personal information isn’t disclosed while you enjoy all the basic benefits of ChatGPT.

AnonChatGPT only allows one conversation at a time and past responses aren't stored. It’s a perfect ChatGPT alternative if you want to ask questions that you may not want to be linked to your account or identity.

• Creator: AnonChatGPT
• Pricing: Free

Try AnonChatGPT

YouChat is your personalized and conversational AI assistant.

It's a ChatGPT clone created by the You.com search engine to converse like humans and provide responses when prompted. The tool can write emails, translate, summarize text, learn language, write code, and everything else a generic chatbot can do.

YouChat was built on top of the GPT-4 model.

• Creator: You.com
• Pricing: Free. YouPro costs $20 per month

Try YouChat

OpenAI Playground

The OpenAI Playground is a web app and ChatGPT alternative that allows you to interact and experiment with GPT4 and other OpenAI models.

The web app works just like ChatGPT, gives you access to the OpenAI API, and also lets you try out advanced technical functions.

You can choose your preferred AI model, set temperature, add a prompt role for the system, set the number of tokens, and much more.

• Creator: OpenAI
• Pricing: Free to use

Try OpenAI Playground

Poe is Quora’s conversational AI platform that allows you to explore and interact with various AI chatbots and models.

The platform's name stands for Platform for Open Exploration and it connects to several generative AI models including ChatGPT, GPT-4, Claude Plus, Sage, Dragonfly, and more.

With Poe, you have a universal messaging app to converse with several AI models brought under one house. The platform is available as an app on the iOS AppStore.

• Creator: Quora
• Pricing: Free. Poe Pro for $19.99 per month

Amazon CodeWhisperer

Amazon CodeWhisperer is Amazon’s entry into the AI chatbot space.

It’s a machine learning-powered code generator that pinpoints errors and recommends code in real time.

While writing coding, the ChatGPT alternative automatically generates suggestions based on your existing code inputs and comments.

This suggestion could be a small one-line comment to a full block of code. Amazon CodeWhisperer is an efficient way for developers to code faster and debug code problems.

• Creator: Amazon
• Pricing: Free for individual use

Try Amazon CodeWhisperer

Copy.ai is an AI writing tool that helps you create headlines, web copy, long essays, and blog articles.

It was built on top of OpenAI’s GPT-3 model and generates content in response to prompts through a smooth interface.

Whether you want to write social media descriptions or generate blog ideas, you’ll find Copy.ai useful. Its free version offers limited usage while the paid plan comes with unlimited features, unlimited brand voices, and access to GPT-4 access.

• Creator: Copy.ai
• Pricing: Free plan (below 2,000 words)

Try Copy.ai

Used by mostly academic researchers and students, Elicit is an AI research assistant that’s built on various LLMs such as GPT-3.

The platform will help you speed up your research process and get relevant citations.

The platform allows you to get accurate, precise, and quick access to information from dependable research sources.

You can simply type a query into the AI assistant and the tool will immediately return a summarized version of the highest-rated sources that match your query.

• Creator: Elicit
• Pricing: Free. Paid plan starts from $10 per month

Learnt.ai is an AI assistant that helps education professionals do their work efficiently and save time.

The platform relies on the GPT AI model to help education professionals generate lesson plans, homework tasks, exam questions, and other types of educational content.

It’s a ChatGPT alternative tailored for educational projects of different levels. To use Learnt.ai, choose a tool suited to what you want to do, prompt the AI assistant, and wait for a response.

• Creator: Learnt.ai
• Pricing: Free to use. Paid plan costs $9 - $99 monthly

Try Learnt.ai

Grok is a generative AI chatbot for understanding the universe. xAI, Elon Musk’s owned startup, developed and rolled out the chatbot as a response to the unprecedented rise of ChatGPT.

The AI chatbot relies on its own LLM called Grok-1 , which was partly trained with content available on X (formerly Twitter) along with other data sources.

In March 2024, Grok’s AI model was open-sourced and the chatbot functionality is now available to use as a standalone app and on X (only for all premium X subscribers).

• Creator: xAI
• Pricing: Available for premium X subscribers

Flawlessly.ai

Flawlessly.ai is an easy-to-use, free ChatGPT alternative for fixing spelling, grammar, tone, and punctuation errors fast.

Essentially, the AI tool makes your writing flawless. To try it out, head to the website, paste the content you want to correct in the box provided, choose a writing tone/style (formal, semi-formal, or conversational), and hit the Correct This button. Flawlessly.ai will return a grammatically correct version of your content in a few seconds.

It’s a must-have writing tool for anyone who struggles with the English Language and needs a free AI grammar checker by the side.

• Creator: Flawlessly.ai

Try Flawlessly.ai

Elsa Speaks

Elsa Speaks is an AI-powered language learning, voice recognition, and translation platform.

The web platform assists you with learning how to speak English and translating different languages. Its underlying AI model was trained using voice recordings from people speaking English with different accents.

Elsa Speaks can recognize vocal patterns (even if you’re not a native English speaker), examine user speech, and generate bite-sized custom tasks that you can understand with ease.

It’s the perfect ChatGPT alternative if your primary use case is language learning and translation.

• Creator: Elsa Speaks
• Pricing: Free Trial. Pro Plan starting at $11.99 per month

Try Elsa Speaks

iAsk AI is a free-to-use advanced AI search engine that instantly provides detailed and accurate answers to your queries in a natural language without any data storage.

Built on similar models and technologies to ChatGPT, the AI search engine makes a great alternative to OpenAI’s chatbot.

Simply type a prompt into the box, choose the website source category (Academic, Forum, News, Wiki, etc) that you want to use, set the length of result you want, and the search engine will generate factual results for you.

iAsk AI has been trained exclusively using reliable and authoritative books, news, literature, and website sources so it can provide objective, accurate, and bias-free responses.

• Creator: AI Search Inc

Try iAsk AI

Other ChatGPT Alternatives You Should Know

Depending on your use case, here are some other AI-powered conversational chatbots you should be familiar with:

• SocialBu’s AI Content Generator
• Socratic (by Google)
• HuggingFace
• AdCreative.ai
• LaMDA (Language Model for Dialog Applications)
• Ink for All
• Megatron-Turing Natural Language Generation
• LLaMA by Meta

FAQs on ChatGPT Alternatives

What is the best alternative to chatgpt.

The best alternative to ChatGPT completely depends on your specific needs. However, our recommended general-purpose AI chatbots are Google’s Gemini, Microsoft’s Bing AI, and Monica–our all-in-one AI assistant!

What are some drawbacks of ChatGPT?

Despite its rapid growth in userbase, powerful generative capabilities, and the launch of Custom GPTs, ChatGPT still has certain downsides you may have noticed. Some of these drawbacks include unavailability during peak periods (when using the free version), slow response time, no advanced customization options, limited UI capabilities, and hallucinations.

Final Words

Ready to explore the unique potential and capabilities of these conversational AI chatbots mentioned in this post offer?

Take any of them for a spin, based on your needs!

Subscribe to Monica Blog

Navy chief charged with espionage used Google as source for ‘white papers,’ defense argues

SAN DIEGO (KGTV) — A military judge must decide if a Navy chief was simply writing research papers for money or if he committed espionage and betrayed the United States.

Attorneys for Chief Petty Officer Fire Controlman Bryce Pedicini argued Tuesday the sailor simply copied and pasted information he found on Google.

Pedicini sent a total of eight “white papers” to someone posing a Japanese researcher on Facebook. That person offered to pay Pedicini cash in exchange for military information starting in 2022.

“It was a very stupid thing,” Pedicini told NCIS investigators when he was arrested abroad in Japan.

Navy Chief Bryce Pedicini has no comment for me as he leaves the close of his espionage trial on Naval Base San Diego. Prosecutors allege he sold sensitive military info to a stranger on the internet for cash. Defense says the case has holes & evidence is missing. ⁦ @10News ⁩ pic.twitter.com/CajDNY2meE — Austin Grabish (@AustinGrabish) April 17, 2024

His attorneys told judge Cmdr. Andrea Lockhart, who is overseeing the espionage trial on Naval Base San Diego, that he fully cooperated with investigators, gave passwords to his devices, and outlined his relationship with the supposed researcher.

National security concerns

The prosecution has said the researcher was an intelligence officer working for a foreign government. The person is only identified as “individual one” in court.

Due to national security concerns, parts of the case, including information about the spy, are being heard in a classified closed session not accessible by the public.

Pedicini’s attorneys argue evidence is missing from the case and said the prosecution hasn’t proven its case beyond a reasonable doubt.

The government fought back Tuesday afternoon and maintains Pedicini was motivated by greed and “cashed out” on his security clearance betraying his fellow sailors.

Court heard while stationed in Japan, Pedicini took photos inside a secure room of classified computers. He sent those images to the intelligence officer through a website the spy set up.

Classified information on China, Russia

In one instance, the Navy chief said, “an amount closer to $5,000 would be decent,” when negotiating how much he’d get paid for the military information.

Pedicini used a burner phone and the Telegram encrypted messaging app to cover his tracks, prosecutors allege.

He pleaded guilty to a single charge stemming from taking a personal phone into a secure room last week. It’s the more serious espionage and communicating defense information charges he is fighting.

The prosecution has not identified the country the U.S. believes the intelligence officer was working for.

Court previously heard some of the information Pedicini sent included details on Chinese and Russian threats as well as a classified document relating to a ballistic missile’s system.

The judge will begin her deliberations on Wednesday and deliver a verdict at a later time.

"This is a complicated case," Lockhart said Tuesday.

Sign up for the Breaking News Newsletter and receive up to date information.

Now signed up to receive the breaking news newsletter..

Free Books for Kids!

• International edition
• Australia edition
• Europe edition

Is writing down my rage the secret to resolving it?

New research reveals that listing your grievances on a piece of paper, then throwing them away may make you less angry. So I gave it a try …

A lifetime enveloped in a benign, insulating cloud of oestrogen left me ill-prepared to be this nakedly, shockingly angry as it ebbs away in perimenopause. It is occasionally exhilarating, but mainly awful, being furious about so many things: the government, contradictory dental advice, inaction on climate breakdown, whatever cat keeps defecating at my back door. I exist at an exhausting, irrational rolling simmer that periodically comes to a head with me inappropriately venting, realising I’m being unreasonable, shamefacedly having a word with myself, then getting cross again.

Help may be at hand, however, according to research from Japan , which suggests that writing your grievances on paper then throwing it away may make you less angry. Study participants were deliberately angered by researchers criticising their work and adding gratuitous insulting comments. Participants then wrote down how they felt and either threw the paper away, shredded it or kept it. The ones who disposed of the paper “completely eliminated their anger”.

Although participants were mostly in their – surely less irate? – early 20s, I had to test if it worked on midlife rage too. Dissipating my Boris Johnson or Rwanda policy rage seemed an impossibly big ask, but I wrote them down, plus some pettier, momentary furies: “Why am I the only one who deals with expired hummus?”; “Man in gravel driveway methodically spraying tiny, hopeful dandelion shoots”; “Malfunctioning laptop fan”; and “Towels piled up wet”. All classics.

I found the act of crumpling or ripping (I don’t have a shredder) very physically satisfying – a tiny haptic catharsis. But, examining my feelings afterwards, I was as tooth-grindingly angry about Johnson’s existence as ever. The smaller irritations mostly melted away, but I think the mere act of committing my hummus chuntering to paper was enough to show me how ridiculous I was in quite an anger-deflating way. I’m not sure that disposing of my angry scrawls added anything. Perhaps what would really help my anger management is a running grievance list, whereby I write about everything and everyone I’m cross with. What could possibly go wrong?

• Health & wellbeing
• Mental health

Most viewed