ChatGPT: Our research demonstrates that AI is capable of generating journal-worthy academic papers, even though some journals prohibit its use

共有用URL https://everevo.com/event/78186
開催日程

2023/11/24(金)09:15 ~ 15:15

カレンダーに追加
詳細

Certain eminent academic journal publishers across the globe have prohibited or the use of ChatGPT, an advanced chatbot, by their authors. Publishers are concerned that inaccurate or plagiarized work may find its way into scholarly publications due to the bot's reliance restricted on internet information to generate answers that are exceptionally legible.

Already, a number of scholars have cited the chatbot as a co-author on scholarly articles, and certain publishers have taken measures to prohibit this conduct. However, to an even greater extent, the editor-in-chief of Science, a preeminent scientific journal globally, has strictly prohibited the incorporation of program text into submitted papers.

It is unsurprising that academic publishers are intrigued by the implementation of discussion algorithms. Our recent research, which was published in Finance Research Letters, that ChatGPT could be utilized to compose an academic journal-acceptable finance paper. While the algorithm exhibits suboptimal performance in certain domains, the integration of our own expertise assisted journal reviewers in identifying and addressing the program's shortcomings.

On the contrary, we contend that publishers and researchers ought to regard ChatGPT not as a potential menace but rather as a potentially valuable research assistant – an inexpensive or even complimentary digital assistant.

Our rationale was that if favorable results can be obtained effortlessly through the use of ChatGPT, then perhaps there is an additional step we can take to elevate these favorable results to exceptional ones.

We initially requested that ChatGPT produce the conventional four components of a research study: a research proposal, a literature review (an assessment of prior scholarly pertaining to the same subject matter), a dataset, and recommendations for experimental procedures and analysis. solely delineated the overarching topic and stipulated that the result should meet the criteria for publication in a reputable finance journal.

This was the initial of ChatGPT that we selected. For version two, we pasted just under two hundred implementation abstracts (summaries) of pertinent, previously conducted research studies into the ChatGPT window.

Subsequently, we requested that the program incorporate these factors into the design of its four research phases. In the final version, version three, we incorporated "domain expertise"—scholarly researcher input. We provided feedback on the computer program's generated answers and offered recommendations for enhancements of connections. We effectively combined our knowledge and skills with those of ChatGPT.

Then, we requested that one variation of how ChatGPT can be utilized to generate an academic study be evaluated by a panel of 32 reviewers each. In order for the output to qualify for publication in a "good" academic finance journal, reviewers were tasked with determining whether it was adequately exhaustive, credible, and novel in its contribution.

Expert reviewers deemed all of these studies to be typically acceptable; this was the most important takeaway. It is truly remarkable that a chatbot was considered competent in producing high-caliber ideas for academic research. This situation gives rise to essential inquiries concerning the definition of creativity and the rightful proprietorship of innovative concepts—questions that remain unanswered at present.

Strengths and weaknesses of each individual

Additionally, the results underscore certain possible merits and demerits of ChatGPT. It was observed that distinct research sections received varying ratings. In general, both the research concept and the dataset received excellent marks. The literature reviews and testing suggestions received a modest, somewhat satisfactory , evaluation.

Our hunch is that ChatGPT is especially adept at connecting a collection of external texts (the essence of a research concept) or modifying easily accessible sections of a single document (the data summary is an example of an easily identifiable "text chunk" in the majority of research studies).

A relative deficiency of the platform became evident when the task became more intricate, specifically when the conceptual process consisted of an excessive number of stages. Typically, literature evaluations and testing are classified as such. Generally, ChatGPT performed well at a subset of these tasks but not all of them. It appears that this has been noted by the reviewers.

Nevertheless, these constraints were surmounted in our most sophisticated iteration (version three), through collaboration with ChatGPT to produce satisfactory results. In subsequent reviews, all sections of the advanced research study received high marks, indicating that academic researchers continue to play an important role .

Moral ramifications

ChatGPT functions as an instrument. Our study demonstrated that it is possible to generate a respectable finance research study with some caution. It produces plausible work even in the absence of caution.

There are obvious ethical implications to this. In academia, research integrity is already a pressing concern, and websites like RetractionWatch disseminate a constant torrent of fabricated, plagiarized, and simply incorrect research studies. Could ChatGPT prolong this issue further?

Potentially, is the succinct response. However, it is impossible to return the genie to its receptacle. Rapidly, the technology will also continue to advance. Concerning precisely how we might recognize and regulate the function of ChatGPT in research is a matter for a different time. Nevertheless, our findings are also applicable in this regard: by demonstrating that the ChatGPT study version incorporating the expertise of researchers is superior, we demonstrate that human researchers' input remains crucial for credible research.

At present, it is our belief that researchers ought to regard ChatGPT as a tool rather than a menace. Emerging economy researchers, graduate students, and early-career researchers, who typically lack the financial means for conventional (human) research assistance, may find This to be an especially useful resource. ChatGPT (and comparable applications) may potentially contribute to the democratization of the research process.

However, scholars must be cognizant of the prohibition against utilizing it in the synthesis of scholarly articles. It is evident that broad perspectives exist regarding this technology; therefore, it will require cautious application.