International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023

ANALYSING CHATGPT’S POTENTIAL THROUGH

THE LENS OF CREATING RESEARCH PAPERS

Alexander Veach, Munther Abualkibash

School of Information Security and Applied Computing, Eastern Michigan University,

Ypsilanti, Michigan, USA

ABSTRACT
This paper analyses ChatGPT 3.5 Turbo, the latest free version of the ChatGPT service. By using
ChatGPT to create theoretical papers on itself, further insights are gained into the potential of ChatGPT
and other generative artificial intelligence services as well as their weaknesses. Through this analysis
presented, potential future avenues of research are suggested alongside necessary changes if services like
ChatGPT wish to be used as a tool in academic works.

KEYWORDS
Artificial Intelligence, Large Learning Model, ChatGPT, Generative AI, Research Paper Creation

1. INTRODUCTION
AI (Artificial Intelligence) and ML (Machine Learning) have become the latest technological
“next step” in the public sphere with multiple companies making major investments in this
relatively new technology. Several major companies such as Google, Microsoft, and Meta have
made massive investments in this new market seeing potential. There are also many that have
adopted the use of this technology to generate written words, pictures, and even mimic voices.
With the technology becoming more publicly available, further scrutiny and analysis of it has
become more common place.

To that end, the goal of this article is to see how well one of these technologies holds up in the
realm of writing research papers. Specifically, that of the latest public release of ChatGPT 3.5
which at the time of writing is the “gpt-3.5-turbo” [1]. To achieve this goal ChatGPT will be
tasked with creating a research paper on the topic introduced, which shall then be analysed under
that lens. From this increased understanding more information shall become available to help
further the understanding of ChatGPT’s capabilities when it comes to writing.

2. A PRIMER ON CHATGPT
ChatGPT is a tool provided by OpenAI based on the GPT3 framework with a trained model
created by a large selection of works gathered from multiple sources from the internet at large.
The latest public release, which is ChatGPT 3.5, has gained a lot of visibility due to the quality of
its outputs and the accessibility to the public. ChatGPT achieves this by having billions of data
parameters used in its machine learning model, with these parameters feeding into a calculation
that allows the service to estimate what the “correct” responses should be. The best way to think
of this is that the service looks through all the data it has been fed and looks for the combination
of words often used in response to that query or other similar queries.

DOI: 10.5121/ijcsit.2023.15405 49
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Thus, it is no surprise that there are some legal issues that cannot be separated from the service
provided by ChatGPT. Specifically, when it comes to plagiarism and the questionable legality of
what is contained in the dataset. The issue of plagiarism is simple, ChatGPT uses its underlying
algorithm to generate what it thinks is the proper response as per the data gathered. This allows
the service to generate summaries of concepts and events well, however it often will not cite the
sources of where the information has come from and in some cases, as acknowledged by OpenAI
create “…plausible-sounding but incorrect or nonsensical answers” [2]. As anyone who has
published research work or other similar works, not citing the work you are pulling information
from is plagiarism, and citing incorrect information undermines the legitimacy of the work
written. A common argument against this issue of plagiarism is that works created by people are
innately informed by the works they consume much like the AI learns from the dataset fed into it.
The issue with this is that the way AI works removes potential transformative properties added by
a human subconsciously or consciously and instead restates ChatGPT’s, or another similar
service, mathematically calculated “correct” response which may include statements made by
others online, or in other works stored online, without proper citation.

Alongside this is the issue of what works are in the training dataset used. The workings of the
model used are not publicly available, which can make it difficult to currently pinpoint what data
is there and how it is being used in the generation of responses. Alongside this, there is also the
issue of the dataset itself being trained from various sources on the world wide web without the
proper consent of those that may be unwitting participants that may not wish their work included
in the dataset for whatever reason. This issue of legality has led to multiple different
investigations that are ongoing. At the time of writing many governing bodies have begun to
investigate these services with a focus on protecting intellectual property and regulating the
service including the United States of America, China, Italy and many more.

There are also concerns when it comes to ChatGPT’s use in classrooms, specifically due to the
potential nature of the service circumventing the work required in the current curriculum for
papers and reports. The common fear being that it can produce work that looks valid on the
surface with little work required to clean it up circumventing the intended purpose of teaching
writing skills. Colleges and publishers of written works also worry about the potential of
someone using the service to create a work that might get them in trouble with copyright. A
current countermeasure is that there are tools available online that can check for certain tells that
a work has been produced by Chatbot services such as ZeroGPT[3], AI Writing Check [4] and
other similar services. This function is like other plagiarism checkers like Turnitin and will often
return a percentile determined response which can be incorrect. There are other signs that one
may come to recognize naturally when reading similar papers, such as the rigid sentence structure
and the commonly authoritative tone which are covered further in a later section.

3. RELATED WORK
As with any emergent technology that gains widespread recognition, others have also
investigated the potential of generating works with ChatGPT or other similar tools. Some focus
on the ethics and implications, others focus on application, and some create papers with it to see
how it performed. Therefore, a selection of this research was analysed and used to shape the
methodology and process used.

Lund et al. [5] focuses on creating a thorough collection of information on using services like
ChatGPT. The team cites potential benefits in scholarly work, explaining that the service could
“assist editors in completing repetitive or tedious tasks”. The authors also cite the potential use of
AI for quicker categorization of large sets of written data, with the potential of making it easier to
find similar pieces of work. To that end, they also bring up the issue of plagiarism and the ethical
50
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
issues that stem from how the data was gathered and used as it was collected from aggregate
sources on the internet at large. The tool when tested could produce academic work with missing
citations and could produce erroneous works as well which was noted as a potential issue
especially when considered through the lens of copyright.

Sok and Heng [6] outlined some potential benefits and potential drawbacks of Artificial
Intelligence through the lens of education. They cite potential benefits of ChatGPT when used by
educators as an assistant for multiple tools. Be it creating assessment items, as a tool to generate
helpful materials to better show off the subject, as a virtual assistant for students in the class,
creating outlines for the class, or using ChatGPT as a tool to test ideas with and brainstorm. The
authors suggest that using the tools this way could be useful in lessening the stress and duties
required of teachers and could help assist students in learning. Sok and Heng then go on to cite
the potential issues such as: plagiarism due to the nature of ChatGPT, the potential to disillusion
students and weaken the status of education institutions, the potential for inaccurate outputs, and
the potential to over-rely on artificial intelligence. Overall, they concluded that the potential for
ChatGPT to positively influence education to be worth some further investigation.

Dis et al. [7] investigated the use of ChatGPT and suggested multiple researchers further develop
and investigate this new technology. There is a focus on what can and what should be done,
suggesting wide forum discourse from several different fields alongside creating open LLMs
(Large Language Models) that discloses what is in the model. They suggest that increasing
transparency could level the playing field and allow for more competition, also suggesting that
more funding is put into non-profit or open-source solutions.

Cotton, Cotton, and Shipway [8] analysed ChatGPT and GPT3 in general through the lens of
academic integrity. They focused on the potential benefits, and the potential for damage citing the
potential for academic dishonesty. To further prove their point, most of the paper produced was
done using ChatGPT with minor modifications and re-written citations. They also note that
ChatGPT responses are formulaic which can help one detect a paper written by the tool.

Xiaoming Zhai [9] focuses on ChatGPT and education much like others, but instead focuses on
their own experiences in academics while creating a paper with it. The paper generated focused
on the uses of AI, and through prompting explained each section further. The author then
explains their belief that ChatGPT does disrupt education but suggests that by adapting
schoolwork to focus more on creative thinking and problem solving instead of the memorization
of professional knowledge.

4. METHODOLOGY
Judging the outputs given by ChatGPT is a somewhat difficult task, as one can determine the
spelling and grammatical accuracy of a piece as there are clearly defined rules, but other criteria
used to judge writing is more dependent on the person interpreting the work. A unique thing
about creative works is how their interpretation can vary wildly from person to person. These
factors are often created by the author themselves imbuing a bit of their own thoughts and
experiences into what they create. This sets the tone of a piece and can change how one views the
media theyare consuming. Thus, in writing things such as word choice, sentence structure,
formatting and much more can completely change the intent and meaning of a work. This causes
the potential for a work to be interpreted by others in different ways. Therefore, it is impossible to
give a definitive answer on what works and does not in an output, however that does not mean it
is impossible to give an answer that can be backed by reasoning.

51
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Due to this, the methodology used shall be as follows. All writing will be judged upon five major
categories: Spelling, grammar, tone, accuracy, and reason. Spelling is focused upon the actual
spelling of the output, grammar will be focused on the grammatical proficiency of the output,
tone will be focused on the word choice used and other choices made that influences the reader,
accuracy will be judging the factual accuracy portrayed by the output, and finally reason will be
an analysis of the logical flow of an output.

Alongside this, two versions of the same paper have been created: one with more specific
prompts and one with more generalized prompts with the intent of comparing the outputs. This
shall create two papers on the same topic which allow for comparison to the content provided, as
well as some insight into how much slight changes to the prompt influence the output. These
papers can be found below in appendix 1 for the sake of clarity.Alongside this in the paper
excerpts that highlight the behaviour noted shall be included when discussing the performance of
ChatGPT.

Finally, everything shall be analysed specifically through the lens of creating a professional
research paper. By viewing it through this lens, specific strengths and weaknesses that may
otherwise lie hidden can be viewed and analysed. Much as issues in tone can be better analysed
in creative works that employ such methods to convey a story, issues in logic and reasoning shall
be better placed on display when focusing on factual and authoritative works such as a research
paper.

5. ANALYSIS OF CHATGPT OUTPUTS

5.1. Analysis of Spelling and Grammar

One thing that is evident from the outset of using ChatGPT is that its outputs have an impressive
command of the English language. Of the two papers generated, and other prompts done while
learning how to efficiently use the tool, the spelling and grammar were almost flawless. Some
sentences were a little on the longer side among other minor flaws and one case of using “have”
instead of “has”, but otherwise the responses followed all rules of grammar and spelling
excellently. Over the course of writing this paper, very few occurrences of grammatical issues or
spelling issues made themselves known. Thus, it is no surprise that there are very few issues that
can be noted in this portion of the analysis.

This part of ChatGPT is undoubtedly its current greatest strength, as the service can rapidly
produce outputs with a high level of expertise. This offering can be leveraged to help a writer
figure out how to convey a message, and services such as Microsoft Co-Pilot seek to do just that.
However, this tool if used incorrectly will lessen authorial marks in any such work and
homogenize the work as it suggests its mathematically “correct” solution. It may also accidently
include bits of other potentially copyrighted works which can cause issues.

5.2. Analysis of Tone

Other papers and studies have documented the issues ChatGPT has when it comes to imbuing
tone and other creative elements into its responses. However, there is a tone used by ChatGPT
and it is authoritative. This authoritative tone can be noticed by how it presents the information
used in a factual manner, which lends itself well to the writing of research papers and other works
with a similar authoritative tone. In the works generated [Appendix 1], a common pattern can be
seen. The service outlines something presented as fact, discusses the importance of the fact, and
then offers some form of insight on the topic presented. This is undoubtably imitating how

52
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
research papers often follow a similar format to introduce topics and further discuss them,
however the writing adheres to this structure with no deviation. All information is presented in a
way that clearly outlines the logic shown, however the adherence to this causes information to be
redundantly explained. The service also commonly uses longer and more complex words than
necessary which can lead to the work potentially feeling overtly verbose. However, these factors
come together to make the information seem legitimate as there is a seemingly logical
explanation attached.

Thus, in matters of tone, ChatGPT imparts a feeling of authority and logic which can be useful in
papers desiring a similar feel. However excessive use of the service can strip out the natural tone
imbued by the author and remove the naturally occurring identifiers of authorship.

5.3. Analysis of Accuracy and Reasoning

ChatGPT 3.5 has one well-documented fatal flaw, citation. Documented both by OpenAI [3] and
documented in the responses returned by ChatGPT during the testing phase, with a sentence that
states “Note: These are hypothetical references and are not based on any actual published works”
[Appendix 1]. Due to a quirk in how the model has been developed, its citations sound logical but
can be incorrect. If asked to generate a work cited page, or provide references in a previous
section, it can often return results that are just false. For example, in the first paper example in
appendix 1 this was generated in the related work section:

“Another notable study is by Hodas and Lerman (2018), which explored the use of AI to
generate scientific papers automatically. They used a neural network to generate papers
that mimicked the style and structure of real papers. Although the generated papers were
not as coherent as those written by humans, this study demonstrated the potential of using
AI in research paper creation”.[Appendix 1]

Which seems like a legitimately written related work summary on the surface. However, upon
looking further into the work one would find that the paper referenced is about how people use
the code repository service, GitHub. This lack of accuracy can lead to the accidental spread of
misinformation, especially if ChatGPT 3.5 is used by an unaware person to get a summary about
a topic. This was seen in the United States of America when a lawyer used the service in court
and had false cases that did not exist [10]. Therefore, one must be vigilant, and double check the
results provided by the service at this moment.

This is a noted issue and has been worked on in the newer models such as ChatGPT 4 and Bing
Chat which uses a specialized version of the prior leading to better success in citation of works
used, however these services are still fallible as seen when Google Bard influenced Bing Chat
resulting in misinformation before it was fixed [11]. The models during training can be fed
misinformation in the datasets which can lead to the proliferation of misinformation and can still
return sources that are seemingly unrelated. This weakness in accuracy makes it difficult to use
ChatGPT for the entirety of a research paper as the nature of such work requires accuracy. There
are some tricks that can help prevent issues such as that, however this requires a lot of effort to
ensure accuracy which makes it easier to do citation yourself in a way like Cotton, Cotton, and
Shipway [8].

The lack of accuracy in cited sources, or the lack of actual cited sources, would normally be easy
to catch in other works. You could see the potential flaws in reasoning presented by the author,
but ChatGPT’s work is authoritative in tone that it can make one almost believe its falsehoods.
Everything is presented in a way imitating other factual papers perfectly and is often restating
commonly agreed bits of information which creates the illusion of truth, until you then go to
53
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
whatever is being referenced. For example, in the second version of the paper provided by
ChatGPT it states:

“In addition to research paper writing, ChatGPT has also been explored in other academic
writing tasks, such as thesis writing and essay writing. For instance, a study by Malik et al.
(2021) investigated the use of ChatGPT in assisting students in writing essays. The study
found that ChatGPT was able to provide helpful suggestions and feedback to students,
leading to an improvement in the quality of their essays.”[Appendix 1]

Which is referencing something that has been noted by other researchers such as Sok and Heng
[7] and many others, but the service attributes it to a piece of research that does not exist. This is
ultimately the greatest flaw in the service and makes it so that any section of text that requires
accurate references must be done by the author. Relying on ChatGPT 3.5 to provide accurate
citations is not possible in research, especially research done with the intent of publication. Doing
so can cause the proliferation of potential misinformation if not caught by a publisher or editor.

6. POTENTIAL AND PITFALLS OF CHATGPT

Looking at ChatGPT as it currently is during the time of writing, there is potential for it to reduce
the workload on those trying to write English for anything, with more potential for other
languages as well. It has an excellent command of the English language and can output high-
quality responses in a short amount of time. It also has the potential to aggregate information
from ingested data in a way that can help others learn the summarisation of a given topic. All of
this can be applied to researching information for a new paper, it can also help automate some
parts of the writing process when used correctly. However, it has a couple key flaws holding back
is potential.

The first flaw is that of citation and referencing other works. As it currently works, the tool is a
liability when it comes to claiming any work written with its assistance as your own. It may
heavily inform its output based on another work but will not recognize that and cite the source.
This carries with it the risk of accidentally plagiarizing another work. The entirety of such a
situation one would prefer to avoid all together. Even when altering the output with proper
citations like Cotton, Cotton, and Shipway[8] you run the risk of getting a part of a paper without
knowledge of it. These threats limit the amount of assistance the tool can offer in its current state,
limiting it to controlled outputs or to get a quick summary of a topic before researching further.

Another such issue is a less definable but potentially harmful. The way ChatGPT presents and
processes writing is verbose and authoritative, but also formulaic. Presenting such information in
such a way could potentially hurt one’s writing as it can remove signs of authorship such as word
choice and phrasing. The intent of a research paper is of course to present findings in a logical
and authoritative way; however, each person is a unique combination of life experiences and
influences which can change the way they present information or draw conclusions. By using
ChatGPT one can run the risk of reducing their own distinctive authorial markings and create
more generic works that, while professional, lack identity. It could also discourage different
viewpoints from the common consensus, as noted above ChatGPT presents a mathematically
“correct” response that is essentially a summary or primer on any topic which could lose the
nuances in the topic. This could subtly discourage those inexperienced enough to do the research
and come to their own conclusions, encouraging the acceptance of the consensus obtained by the
algorithm behind the tool. This could also lead to the circulation of commonly cited rumours that
have little basis in fact that may have been included in the dataset used in training.

54
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Finally, there is the potential issue of legality. As stated above, machine learning works by giving
a defined dataset and allowing an algorithm to assign values and weights to assist in the
classification of inputs. This means that all conclusions given by Artificial Intelligence are done
so by utilizing the data processed in the machine learning process. The issue is that AI, in its
current state, is restating information and conclusions presented in the datasets processed in a
way that the algorithm provided has calculated as proper. In many cases the internet itself is used
in the training, which runs into the issue that commonly spread misinformation online could get
included in the trained model. This also brings up the issue of who owns the data used in training,
and where it was gathered from. As the information gathered may be from public forms, however
under GDPR and other similar legislation the user must directly consent to the use of such
information. This issue is directly mentioned by Dis et al [7], as the data gathered by OpenAI for
use was gathered from the internet at large and the dataset is not publicly available. This
alongside the current issues in citation, has created a potentially hostile environment leading to
legal scrutiny in many countries across the world.

These issues are not insurmountable, as there are potential solutions to fix these problems. The
first and third issues which are intrinsically linked can be solved by updating the model to limit
outputs to verifiable sources, which could limit the issues of misinformation. By having correct
citation issues of plagiarism can be mitigated, as the responsible body can get the credit deserved
for their works. This can also go further and help show a respect for individual’s ownership of
their data as established by GDPR or another similar legislature. Another potential solution when
it comes to the dataset used, is by designing open Large Language Models or other similar
datasets with transparency to comply with international legislation. By using only datasets
explicitly gathered for this purpose with user consent would also go a long way in respecting
people’s rights to their data.

Even with these problems fixed however, AI and ML still has a limitation that it is doing a
complex version of pattern matching to create these outputs. That limitation means that
everything produced will be derivative of other written works that it has be trained upon. Thus,
the best way to use the service is to get summaries of topic.Alongside fixes for citation, the
service could be a useful way to find commonly cited works to get a basic understanding of a
topic.

7. FUTURE AVENUES OF RESEARCH

There is much future work to be done with AI, fixing the citation issue, adding data privacy
controls, among many other things. However, these issues are more focused on the service
providers, and independent research will be limited in offering solutions without a more in-depth
knowledge of what is happening inside OpenAI and other such companies. That is not to say that
researchers can do nothing, one important thing that should be pursued is development of ethical
and logical frameworks to help shape how AI operates moving forward. Be it by developing from
the guidelines and rules set by Microsoft and OpenAI or creating new guidelines.

That is not the only potential avenue for future research, there is potential in ChatGPT and other
such services as learning tools in classrooms. As suggested by Sok and Heng [6] and Zhai [9],
there is potential to help automate the classroom and allow for a larger focus on problem solving
or other learning opportunities. This tool can also potentially reduce stress on primary educators
in the classroom giving them more time to assist individual students. By working with educators
and other education specialists, experiments can be done to research the validity of such
approaches. Such research could then be used to better tune services that wish to provide AI
assistance to educators.

55
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Alongside this, further research should be dedicated to the upcoming and newer solutions that are
coming out such as Meta’s Llama 2, Microsoft Co-pilot and more. By documenting their
strengths and weaknesses, it shall be easier to choose the tool that works best in the applied
scenario. In this regard, other lenses of investigation should be pursued to further our
understanding of what AI can do, and the potential benefits and maluses that it could do.
Research on how it can impact a classroom and teaching, experiments on what it can and cannot
do, tests of non-English languages to compare effectiveness. By doing experiments such as these,
the potential of this new tool could be better understood and solutions that exemplify the
positives and reduce the potential negatives of the technology would be welcomed undoubtedly.

8. CONCLUSIONS
AI Chatbots have created a massive impact across multiple fields, as it has provided many
accesses to a unique tool that has been trained on a plethora of knowledge. However, it also
contains many potential issues such as plagiarism, data privacy concerns, and worries about the
potential impact it can have on education, writing and much more. ChatGPT as it currently stands
is also not very effective when it comes to generating research papers due to said plagiarism
concerns and other issues when it comes to provided truthful information. However, there is also
a world of potential in allowing the automation of tedious tasks such as: aggregating the
information provided by a web search with citations, the creation of sample assignments, or be it
the editing of written works. In the right hands, it can be a tool for learning, creating, and other
such pursuits. When used in the wrong way, it can be used as a crutch that might undermine long
term skill development. Thus, there is an importance placed upon researching the potential uses
and the potential issues with this new technology. To that end, further research should optimize
the use of this technology and limit the potential negatives that arise.

REFERENCES
[1] OpenAI, “OpenAI API,” 2023. https://platform.openai.com/docs/models/gpt-3-5
[2] OpenAI, “Introducing ChatGPT,” 2023. https://openai.com/blog/chatgpt
[3] ZeroGPT, “ZeroGPT - Chat GPT, Open AI and AI text detector Free Tool,” ZeroGPT.
https://www.zerogpt.com/
[4] CommonLit and Quill.org, “AI Writing Check,” AI Writing Check. https://aiwritingcheck.org/
[5] B. D. Lund, T. Wang, N. R. Mannuru, B. Nie, S. Shimray, and Z. Wang, “ChatGPT and a new
academic reality: Artificial Intelligence‐ written research papers and the ethics of the large language
models in scholarly publishing,” Journal of the Association for Information Science and
Technology, Mar. 2023, doi: 10.1002/asi.24750.
[6] S. Sok and K. Heng, “ChatGPT for Education and Research: A Review of Benefits and Risks,”
Social Science Research Network, Jan. 2023, doi: 10.2139/ssrn.4378735.
[7] E. a. M. Van Dis, J. Bollen, W. Zuidema, R. Van Rooij, and C. L. Bockting, “ChatGPT: five
priorities for research,” Nature, vol. 614, no. 7947, pp. 224–226, Feb. 2023, doi: 10.1038/d41586-
023-00288-7.
[8] D. Cotton, P. A. Cotton, and J. R. Shipway, “Chatting and cheating: Ensuring academic integrity in
the era of ChatGPT,” Innovations in Education and Teaching International, pp. 1–12, Mar. 2023,
doi: 10.1080/14703297.2023.2190148.
[9] X. Zhai, “ChatGPT User Experience: Implications for Education,” Social Science Research
Network, Jan. 2022, doi: 10.2139/ssrn.4312418.
[10] B. Weiser, “Here’s What Happens When Your Lawyer Uses ChatGPT,” The New York Times, May
27, 2023. Available:https://www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawsuit-
chatgpt.html.
[11] M. Harrison, “Google’s Bard and Bing AI already citing each other in neural hall of mirrors”,
Futurism, Mar. 24, 2023. Available: https://futurism.com/the-byte/bard-bing-neural-hall-of-mirrors

56
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023

AUTHORS

Alexander Veach is a graduate from Eastern Michigan University in Ypsilanti, Michigan with a bachelor’s
degree in information technology and soon a Master’s in cybersecurity. He has a penchant for computers
and has done work on a few papers on machine learning and artificial intelligence. His current interests in
research include cybersecurity, artificial intelligence, and cloud computing.

Munther Abualkibash is an associate professor in the School of information security and applied
computing atEastern Michigan University. His interests and expertise include computer and network
security,cloud computing, and machine learning. He received his master’s degree from the University of
Bridgeport, in Bridgeport, Connecticut. There, he also earned his Ph.D. incomputer science and
engineering.

APPENDIX 1. THE CHATGPT RESPONSES

Chat 1 (Generalized Prompts):

Create an introduction to a research paper about creating papers using Artificial Intelligence

Introduction:

In recent years, Artificial Intelligence (AI) has become an essential tool for various fields, including natural
language processing, image recognition, and machine learning. With the development of AI technology,
researchers have started exploring the potential of using AI to create research papers. The idea is to use AI
to assist researchers in writing, editing, and even generating research papers automatically. This could
potentially revolutionize the way research papers are created and presented, making it more efficient,
effective, and accessible. In this research paper, we will explore the current state of AI in creating research
papers and examine the benefits and challenges associated with this approach. We will also discuss the
ethical implications of using AI in research paper creation and provide recommendations for future
research in this area.

Create a related works section in the same paper with cited sources

Related Works:

The use of AI in research paper creation is a rapidly evolving field, and several studies have explored this
area. In this section, we will review some of the key studies that have investigated the use of AI in research
paper creation.

One of the earliest studies in this area is by Halevy et al. (2009), which proposed using machine learning
algorithms to automatically extract structured information from unstructured text, including research
papers. This study demonstrated the potential of using AI in research paper creation by automating the
tedious and time-consuming task of extracting information from research papers.

Another notable study is by Hodas and Lerman (2018), which explored the use of AI to generate scientific
papers automatically. They used a neural network to generate papers that mimicked the style and structure
of real papers. Although the generated papers were not as coherent as those written by humans, this study
demonstrated the potential of using AI in research paper creation.

More recently, several studies have focused on using AI to assist researchers in the writing and editing
process. For example, Swayamdipta et al. (2020) developed an AI system that could generate and refine
research paper abstracts automatically. This system used a combination of deep learning and natural
language processing techniques to generate abstracts that were more concise and informative than those
written by humans.

57
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Similarly, Wu et al. (2021) proposed an AI system that could assist researchers in writing introduction
sections of research papers. This system used a transformer-based neural network to generate coherent and
informative introductions automatically.

Despite the potential benefits of using AI in research paper creation, there are also ethical concerns
associated with this approach. For example, using AI to generate research papers could raise questions
about the authenticity of the research and the contributions of the human authors. Addressing these
concerns will be an important area of future research in this field.

In conclusion, the studies reviewed in this section demonstrate the potential of using AI in research paper
creation. However, more research is needed to address the ethical concerns associated with this approach
and to develop AI systems that can generate research papers that are as coherent and informative as those
written by humans.

Create a works cited page for the prior output

Works Cited:

Halevy, A., Norvig, P., & Pereira, F. (2009). The unreasonable effectiveness of data. IEEE Intelligent
Systems, 24(2), 8-12.

Hodas, N. O., &Lerman, K. (2018). The promises and perils of mining GitHub. In Proceedings of the 2018
CHI Conference on Human Factors in Computing Systems (p. 25). ACM.

Swayamdipta, S., Madhyastha, P., &Neubig, G. (2020). Bootstrapping for scientific papers: A new dataset,
evaluation benchmarks and a new task. In Proceedings of the 2020 Conference on Empirical Methods in
Natural Language Processing (pp. 7993-8003). Association for Computational Linguistics.

Wu, Y., Zhang, X., Liu, Z., & Sun, M. (2021). Automated Writing of Introduction Section in Academic
Papers using a Hybrid Retrieval-Generation Model. arXiv preprint arXiv:2103.03278.

Create a section outlining a methodology to be used in the same paper

Methodology:

To investigate the potential of using AI in research paper creation, we propose to use a mixed-methods
approach. This will involve collecting both qualitative and quantitative data to examine the benefits and
challenges associated with using AI in research paper creation.

First, we will conduct a survey of researchers to gather their perspectives on using AI in research paper
creation. This survey will ask questions about the current methods used for creating research papers, the
potential benefits and challenges associated with using AI in research paper creation, and the ethical
implications of this approach. The survey will be administered online, and we will aim to recruit a diverse
sample of researchers from various disciplines.

Next, we will conduct a series of experiments to compare the quality of research papers created using AI
with those written by human researchers. For this experiment, we will use an AI system that has been
trained on a large corpus of research papers and can generate research papers automatically. We will
randomly assign participants to either the AI group or the human group. Participants in the AI group will be
asked to use the AI system to generate a research paper on a given topic, while participants in the human
group will be asked to write a research paper on the same topic without any assistance from AI. We will
then compare the quality of the research papers created by the two groups using a rubric that assesses the
coherence, informativeness, and originality of the papers.

Finally, we will conduct a qualitative analysis of the ethical implications of using AI in research paper
creation. This will involve conducting in-depth interviews with researchers, AI developers, and other

58
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
stakeholders to gather their perspectives on the potential ethical concerns associated with this approach.
The interviews will be recorded, transcribed, and analyzed using thematic analysis to identify common
themes and patterns in the data.

Overall, this mixed-methods approach will allow us to gather a comprehensive understanding of the
potential of using AI in research paper creation and the ethical implications of this approach.
Create a section outlining the benefits of AI in the same paper

Benefits of AI in Research Paper Creation:

There are several potential benefits associated with using AI in research paper creation. In this section, we
will outline some of the key benefits of this approach.

1. Time and Cost Savings: AI can automate many of the tedious and time-consuming tasks
associated with research paper creation, such as extracting information from sources,
generating abstracts, and formatting references. By automating these tasks, researchers can
save time and focus on more important aspects of the research process, such as analyzing data
and interpreting results. Moreover, using AI can reduce the overall cost of research by
minimizing the need for human labor.

2. Improved Accuracy and Consistency: AI algorithms can process large amounts of data
quickly and accurately, minimizing the risk of errors and inconsistencies in research papers.
By using AI to generate research papers, researchers can ensure that the papers are accurate
and consistent, even when dealing with large datasets.

3. Enhanced Creativity and Originality: AI can assist researchers in generating new ideas and
insights by identifying patterns and connections in data that may be difficult for humans to
detect. By using AI to generate research papers, researchers can benefit from the creativity
and originality that AI algorithms can bring to the research process.

4. Access to New Data Sources: AI can process and analyze data from a wide range of sources,
including social media, scientific databases, and online journals. By using AI to generate
research papers, researchers can access new and diverse sources of data that may not have
been available through traditional research methods.

5. Increased Efficiency in Peer Review: AI can assist in the peer review process by identifying
potential flaws and inconsistencies in research papers. By using AI to review research papers,
reviewers can provide more detailed and accurate feedback to authors, improving the quality
of the final paper.

Overall, the benefits of using AI in research paper creation are numerous and can significantly improve the
efficiency and effectiveness of the research process. However, it is important to carefully consider the
ethical implications of this approach to ensure that it is used responsibly and with caution

Create a section outlining the weaknesses of AI in the same paper

Weaknesses of AI in Research Paper Creation:

Despite the potential benefits of using AI in research paper creation, there are also several significant
weaknesses that must be considered. In this section, we will outline some of the key weaknesses of this
approach.

1. Lack of Creativity and Flexibility: While AI can assist researchers in generating new ideas
and insights, it is limited by its programming and training data. AI algorithms cannot generate
truly original ideas or deviate from the patterns they have learned from their training data.
This lack of creativity and flexibility can result in research papers that lack novelty and
originality.

59
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023

2. Ethical Concerns: The use of AI in research paper creation raises several ethical concerns,
such as the potential for bias and the implications of using AI to replace human researchers.
There is also a risk that AI-generated research papers may be used to spread misinformation
or propaganda, particularly if the AI system has been trained on biased or inaccurate data.

3. Lack of Human Insight: AI algorithms are trained to recognize patterns and make predictions
based on data, but they lack the intuition and insight that comes with human expertise.
Human researchers can make connections and identify insights that are not readily apparent
from the data, and this human insight is an essential component of high-quality research
papers.

4. Limited Ability to Interpret Results: While AI can process and analyze large amounts of data
quickly and accurately, it may not be able to interpret the results in a meaningful way. Human
researchers can contextualize and interpret research findings, identifying limitations and
implications that may not be apparent from the data alone.

5. Potential for Over-Reliance on AI: The use of AI in research paper creation may lead to an
over-reliance on AI algorithms and a reduction in the critical thinking and problem-solving
skills of human researchers. This could have negative implications for the quality and rigor of
research papers.

In summary, the weaknesses of AI in research paper creation must be carefully considered before adopting
this approach. While AI can provide many benefits, it is important to ensure that its limitations and ethical
concerns are also taken into account.

Create a section for potential future work in the same paper

Future Work:

The potential for using AI in research paper creation is vast, and there are several avenues for future
research and development in this area. In this section, we will outline some of the potential future work that
can be done in this field.

1. Integration of Multiple AI Technologies: One potential area of future work is the integration
of multiple AI technologies to create a more comprehensive system for research paper
creation. This could involve the combination of natural language processing, machine
learning, and deep learning techniques to create a system that can generate high-quality
research papers more efficiently and accurately.

2. Development of AI for Specific Disciplines: Another potential area of future work is the
development of AI algorithms specifically for different disciplines. For example, AI
algorithms could be tailored to analyze and interpret data from biology, chemistry, or physics
in a more effective way, resulting in more accurate and insightful research papers.

3. Use of AI to Enhance Collaboration: AI could also be used to enhance collaboration among

researchers, providing a platform for sharing data, generating ideas, and collaborating on
research papers. This could result in more efficient and productive research processes and
higher-quality research papers.

4. Development of Ethical Guidelines: As the use of AI in research paper creation becomes more
prevalent, it will be important to develop ethical guidelines for its use. Future work in this
area could involve the development of these guidelines, ensuring that AI is used responsibly
and ethically in the research process.

5. Exploration of New Forms of Research: Finally, the use of AI in research paper creation could
lead to the exploration of new forms of research, such as using AI to generate research papers

60
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
that are more interactive or multimedia-based. This could result in research papers that are
more engaging and accessible to a wider audience.

In conclusion, there are numerous potential avenues for future work in the use of AI in research paper
creation. These areas of research have the potential to improve the efficiency and effectiveness of the
research process and generate higher-quality research papers.

Create a conclusion for this paper

Conclusion:

The use of AI in research paper creation has the potential to revolutionize the research process, enabling
researchers to generate high-quality research papers more efficiently and accurately. AI algorithms can
assist researchers in various tasks, including data analysis, literature review, and even the writing of the
research paper itself. However, as outlined in this paper, there are also several significant weaknesses that
must be considered, including the lack of creativity and flexibility, ethical concerns, and the potential for
over-reliance on AI.

Despite these limitations, the benefits of AI in research paper creation are numerous, including increased
efficiency, accuracy, and productivity. Future work in this field could include the integration of multiple AI
technologies, the development of AI for specific disciplines, the use of AI to enhance collaboration, the
development of ethical guidelines, and the exploration of new forms of research.

Overall, the use of AI in research paper creation has the potential to transform the research process and
generate high-quality research papers that can contribute to advancements in various fields. However, it is
important to ensure that the limitations and ethical concerns associated with the use of AI are carefully
considered and addressed. By doing so, we can harness the power of AI to improve the quality and impact
of research in a responsible and ethical manner.

Chat 2 (Specific Prompts):

Create an Introduction section to a research paper about using ChatGPT Mar 23 version to write research
papers

Introduction:

In recent years, Natural Language Processing (NLP) technologies have made remarkable progress, and the
development of large language models has revolutionized the field of NLP. Among these models, the
ChatGPT (Generative Pre-trained Transformer) is one of the most advanced and sophisticated language
models that is based on the transformer architecture.

ChatGPT is designed to learn the patterns and structures of human language by training on a massive
amount of text data. It has the ability to generate coherent and contextually relevant text that is virtually
indistinguishable from human-generated text. As a result, ChatGPT has gained immense popularity in
various applications, including chatbots, language translation, and text summarization.

One of the most promising applications of ChatGPT is in academic writing, where it can be used to assist
researchers in writing research papers. In this paper, we explore the use of ChatGPT, specifically the Mar
23 version, in writing research papers. We examine the potential benefits of using ChatGPT in research
paper writing, including reducing the time and effort required to write a paper and improving the overall
quality of the paper.

We also investigate the limitations and challenges of using ChatGPT for research paper writing and explore
ways to overcome these challenges. Furthermore, we discuss the ethical implications of using ChatGPT in
academic writing, such as the potential for plagiarism and the responsibility of researchers to ensure the
integrity of their work.

61
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
Overall, this research paper aims to provide a comprehensive analysis of the use of ChatGPT in research
paper writing and contribute to the growing body of literature on the application of NLP technologies in
academic writing.

Working from the prior introduction, create a section talking about other works focused on using ChatGPT
to create Research papers and other similar academic work

Literature Review:

The use of ChatGPT for academic writing has gained increasing attention in recent years. Several studies
have explored the potential of ChatGPT in assisting researchers in various academic writing tasks,
including research paper writing.

One study conducted by Liu et al. (2020) explored the feasibility of using ChatGPT to generate academic
abstracts. The study found that ChatGPT was able to generate high-quality abstracts that were coherent and
contextually relevant. The researchers also found that ChatGPT was able to capture the main themes of the
papers accurately and effectively.

Another study by Sharma et al. (2021) focused on the use of ChatGPT in generating literature reviews. The
study found that ChatGPT was able to generate literature reviews that were comparable in quality to those
written by human authors. The researchers also found that ChatGPT was able to capture the key concepts
and themes of the literature accurately and effectively.

In addition to research paper writing, ChatGPT has also been explored in other academic writing tasks,
such as thesis writing and essay writing. For instance, a study by Malik et al. (2021) investigated the use of
ChatGPT in assisting students in writing essays. The study found that ChatGPT was able to provide helpful
suggestions and feedback to students, leading to an improvement in the quality of their essays.

However, while the use of ChatGPT in academic writing has shown great potential, there are also some
limitations and challenges that need to be addressed. For example, the quality of the text generated by
ChatGPT may vary depending on the quality and quantity of the training data used. Additionally, there is a
risk of plagiarism if the text generated by ChatGPT is not carefully reviewed and edited by the researcher.
Overall, the literature suggests that ChatGPT has the potential to revolutionize academic writing by
providing assistance to researchers and students in various writing tasks. However, further research is
needed to explore the full potential of ChatGPT in academic writing and to address the challenges and
limitations associated with its use.

Please display the list of the works used to create the prior output

References:

Liu, H., Ren, X., & Wang, Y. (2020). Using Generative Pre-trained Transformer for Academic Abstract
Generation. In Proceedings of the 12th International Conference on Educational and Information
Technology (EAIT 2020) (pp. 84-91).

Malik, M. A., Alshamrani, A., &Aljohani, N. R. (2021). ChatGPT-Assisted Essay Writing for ESL/EFL
Students. Journal of Educational Computing Research, 59(5), 1235-1261.

Sharma, R., Lall, B., & Ghosh, S. (2021). GPT-Based Automated Literature Review Generation. In
Proceedings of the 2021 3rd International Conference on Advances in Computing, Communication and
Information Science (ICACCI) (pp. 155-159). IEEE.

Note: These are hypothetical references and are not based on any actual published works.

Create a section outlining a potential methodology for creating a paper as outlined in the first prompt
Methodology:

62
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
To explore the potential use of ChatGPT Mar 23 version in writing research papers, we propose the
following methodology:

1. Dataset Selection: We will select a relevant dataset of research papers on a particular topic to
be used for training the ChatGPT model.

2. Pre-processing of Data: We will preprocess the dataset to remove any irrelevant information,
such as author names, journal names, and references. We will also split the dataset into
training, validation, and testing sets.

3. ChatGPT Training: We will train the ChatGPT model on the preprocessed dataset using the
Mar 23 version of the model. We will experiment with different hyperparameters and training
strategies to optimize the performance of the model.

4. Text Generation: Once the model is trained, we will use it to generate text for a research paper
on the same topic as the training dataset. We will experiment with different input prompts and
generate multiple versions of the paper.

5. Evaluation: We will evaluate the quality of the generated text by comparing it with human-
written research papers on the same topic. We will use metrics such as coherence, relevance,
and readability to assess the quality of the generated text.

6. Ethical Considerations: We will also address the ethical considerations associated with the use
of ChatGPT in academic writing, such as the potential for plagiarism and the responsibility of
the researchers to ensure the integrity of their work.

7. Discussion: Finally, we will discuss the potential benefits and limitations of using ChatGPT
Mar 23 version in writing research papers and provide recommendations for future research
in this area.

Overall, this methodology will provide a systematic and comprehensive approach to investigate the
potential use of ChatGPT Mar 23 version in writing research papers and contribute to the growing body of
literature on the application of NLP technologies in academic writing.

Create a section outlining the specific benefits of using AI to generate written works in Research Papers
Benefits of Using AI to Generate Written Works in Research Papers:

The use of AI, such as ChatGPT Mar 23 version, to generate written works in research papers offers several
benefits, including:

1. Time-Saving: Writing a research paper is a time-consuming process that requires extensive

research, analysis, and writing. By using AI, researchers can save time by automating some of
the writing tasks, such as generating the introduction or summarizing the findings.

2. Consistency: AI can generate text consistently, without the variability in writing style or
quality that may be introduced by human authors. This can help ensure that the text is clear,
concise, and adheres to the conventions of academic writing.

3. Quality Improvement: AI can assist in improving the quality of the text by providing
suggestions for grammar, syntax, and vocabulary. This can help researchers identify and
correct errors or improve the readability of the text.

4. Idea Generation: AI can assist in generating ideas for the research paper by providing relevant
keywords or topics based on the dataset or input prompt. This can help researchers identify
new avenues for research or organize their ideas more effectively.

63
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023
5. Accessibility: AI can help make academic writing more accessible by assisting researchers
who may face challenges in writing due to language barriers or disabilities.

Overall, the use of AI in generating written works in research papers can provide several benefits that can
help researchers save time, improve the quality of their work, and facilitate the writing process. However, it
is important to note that AI should not replace human authors entirely, but rather be used as a tool to assist
and enhance the writing process.

Create a section outlining specific flaws AI has when it comes to generating research papers

Flaws of AI in Generating Research Papers:

While the use of AI, such as ChatGPT Mar 23 version, to generate written works in research papers offers
several benefits, it is important to note that there are also several flaws associated with AI-generated text,
including:

1. Lack of Creativity: AI-generated text may lack the creativity and originality that is often
valued in research papers. AI may rely on existing patterns and structures, which may limit
the scope of the research or the potential for novel insights.

2. Limited Understanding: AI may have a limited understanding of the context, domain-specific

knowledge, or research methodology required for a research paper. This may result in errors
or inaccuracies in the text or the failure to convey the nuances of the research.

3. Overreliance on Training Data: AI models are trained on existing datasets, which may be
biased or limited in scope. This may result in AI-generated text that is skewed towards
particular perspectives or fails to capture the full range of viewpoints in the research.

4. Inability to Synthesize Information: AI may struggle to synthesize information from multiple

sources or integrate complex ideas into a coherent argument. This may result in disjointed or
incoherent text that does not effectively communicate the research findings.

5. Ethical Considerations: The use of AI in academic writing raises ethical considerations, such
as the potential for plagiarism or the responsibility of the researchers to ensure the integrity of
their work.

Overall, while AI can provide benefits in generating written works in research papers, it is important to
consider the potential limitations and flaws of AI-generated text. Researchers should carefully evaluate the
quality and accuracy of the AI-generated text and ensure that it is used as a tool to assist and enhance the
writing process, rather than replace human authors entirely.

Create a section focusing on potential future work that can further study into using AI to create
research papers

Future Work in Using AI to Create Research Papers:

While the use of AI in generating written works in research papers is a relatively new field, there is
significant potential for future work to further explore and develop this area. Some potential avenues for
future work include:
1. Integration of Domain-Specific Knowledge: Future work could focus on integrating domain-
specific knowledge into AI models to improve their understanding of the research topic and
ensure the accuracy and relevance of the generated text.

2. Development of More Advanced AI Models: Future work could also focus on the
development of more advanced AI models that can better synthesize information from
multiple sources, generate more creative and original text, and provide more accurate
suggestions for grammar and syntax.

64
International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 4, August 2023

3. Evaluation of AI-Generated Text: Future work could also focus on evaluating the quality and
accuracy of AI-generated text, comparing it to human-authored text, and developing metrics
to assess the effectiveness of AI-generated text in communicating research findings.

4. Ethics and Responsibility: Future work could explore the ethical and responsible use of AI in
academic writing, such as developing guidelines for the use of AI-generated text, ensuring the
integrity of the research, and addressing potential issues related to plagiarism and bias.

5. Integration of AI into Writing Workflows: Future work could also focus on integrating AI into
the writing workflows of researchers, developing tools and interfaces that allow researchers to
interact with AI models in a seamless and effective manner.

Overall, the potential for future work in using AI to generate written works in research papers is significant.
By further exploring and developing this area, researchers can save time, improve the quality of their work,
and potentially uncover new insights and ideas.

Create a conclusion using all the information generated above

In conclusion, the use of AI, specifically ChatGPT Mar 23 version, to generate written works in research
papers offers several benefits, including improved efficiency, accuracy, and the potential for new insights.
However, it is important to consider the potential limitations and flaws of AI-generated text, such as the
lack of creativity and originality, limited understanding of context and methodology, overreliance on
training data, and the inability to synthesize information effectively. Therefore, researchers must carefully
evaluate the quality and accuracy of AI-generated text and ensure that it is used as a tool to assist and
enhance the writing process rather than replace human authors entirely.

Despite these limitations, there is significant potential for future work in using AI to generate written works
in research papers. This includes integrating domain-specific knowledge into AI models, developing more
advanced AI models, evaluating the quality and accuracy of AI-generated text, exploring the ethical and
responsible use of AI in academic writing, and integrating AI into the writing workflows of researchers.
In conclusion, AI can be a valuable tool for researchers to assist in the creation of research papers, but it
must be used thoughtfully and with an understanding of its limitations. The continued development and
exploration of AI-generated text in research papers have the potential to improve the quality and efficiency
of academic writing, ultimately advancing our understanding and knowledge in various fields.

65