How researchers can use ChatGPT to write academic articles

ChatGPT can be a powerful tool for academic writing. It can assist in quickly sifting through vast amounts of information, generate initial summaries of research, and even help to write and edit academic articles. In this post, we provide some initial suggestions for how academics can make best use of ChatGPT and related tools.

Streamlining the search for relevant literature
Searching for articles using Google Scholar and other search engines often requires the laborious tweaking of terms and Boolean operators to unearth the most relevant research. ChatGPT can help to generate lists of academic sources based on the relevant prompts and parameters you provide. It also allows for a more natural interaction when refining your search, helping you to hone in on what you’re looking for through a continuing dialogue with the model.  
Asking ChatGPT to list academic sources and URLs relevant to your chosen topic can save time and effort when conducting literature reviews. However, it's important to remember that this is only a tool and not a replacement for human expertise. Researchers should always conduct their own assessment of the available literature, independent of ChatGPT, and verify the accuracy and validity of the sources provided.

Interacting with resources
Similar to ChatGPT, ChatPDF and Filechat can help to explore documents using AI. With these tools, researchers can select a specific article or report and ask the AI questions about the content of the document such as, "What is the main argument of the article?", "What evidence does the article provide with respect to X?". By using AI to interact with selected documents, researchers can save time and effort in digesting complex information, whilst still maintaining control over the original source of information. This may help us to ask more tailored questions of existing research to improve our understanding of the evidence base.

Quick Outlines and Summaries
The most common use of ChatGPT in academic writing so far has been to generate short summaries that provide an initial overview of a specific topic or issue. These summaries can be particularly helpful for researchers who want to sketch out a paragraph or lay down a road map on which to build.
A similar tool designed to help researchers make sense of vast amounts of academic material is Consensus, which draws information directly from scientific research to provide users with evidence-based answers to their queries. It relies on published, peer-reviewed articles only, as claimed by the company. It aims to generate short summaries of the literature and to inform users of whether or not there is an agreed ‘consensus’ on a specific topic, adding to the growing list of AI tools useful to academics.
​
Editing
When used correctly, ChatGPT can save hours of tedious redrafting and editing. It can help to suggest alternative phrasing or word choices, standardise styles when collaborating with others, and even generate new ideas for structuring and organising ideas.
Researchers can input original text into ChatGPT and it can provide an edited version depending on the defined prompts. This can be particularly helpful when trying to communicate complex ideas to a non-specialist audience. ChatGPT can suggest changes to make text more accessible. For example, ChatGPT can help to convert text from British to American English, change the passive voice to the active voice, or edit the style of text to meet the requirements of a target journal or platform.

Risks
Despite ChatGPT’s potential to assist academic writing, the following risks of large language models must be considered:
Accuracy. A big problem is that ChatGPT does not always produce the most accurate or reliable information, which could lead to fundamental errors in academic writing. For example, many people are currently using the free version of ChatGPT, based on GPT-3.5, which was trained on text data up to September 2021, so it may not have the latest knowledge relevant to a specific field. Also, when ChatGPT receives a question for which it doesn’t have a clear answer, it relies on its training data to generate the most plausible response. This can lead to some pretty audacious responses, sometimes giving the impression that it’s simply winging it!

Integrity. Universities, industry leaders, and policy makers across the world are currently scrambling around to develop guidance on the ethical implications of AI. The use of ChatGPT in academic writing raises important questions around plagiarism and the integrity of research. Researchers should ensure that ChatGPT is used only as a helpful tool to assist in research activities, rather than relying on it as a primary source of information.

Data protection. Another big concern is that when interacting with ChatGPT, users may inadvertently share private information or data protected by the ethical requirements of their research. This information could potentially be stored and used for other purposes without the user's knowledge or consent. Data generated by user interactions with ChatGPT could be used to train other AI models or sold to third parties for profit, without the user's knowledge or consent.
​
Watch this space
As we consider the potential implications of ChatGPT and future language models, it’s clear that these tools offer a wealth of possibilities for academics. From improving the clarity and precision of writing to generating new ideas and insights, AI has the potential to transform the way we conduct and communicate our research. However, it’s important to be mindful of the ethical implications of using these tools, as well as the potential impact on human creativity and critical thinking.