How do researchers use generative AI? How should they use it? This depends in part on the policies of the institutions they are attached to but these policies are changing as researchers come up with new ideas.
For this advice I am again grateful to Anna Clemens, writing coach and founder of the Researchers’ Writing Academy. I have embellished on her ideas with some thoughts from my own experience.
Use gen AI as a tool only
Don’t just copy and paste the generated text. This is the number one point I cannot emphasise enough. I have written elsewhere on my blog about how to identify AI generated academic writing. This is not difficult. Also, rather than taking it as the foundation of your research outline, rely on it to inform your ideas only. You can think of more points. That is what will make your research original.
Critically analyse whether the AI is actually saving you time
It is worthwhile evaluating whether the time you save is worth the (lack of) quality in what is being generated. Consider balancing the effort you make with the value of the writing. This might be hard for researchers working in a second or third language. In these cases it is not so easy to identify paraphrasing as repetition or to notice if a sentence really has a true meaning. It might just be a smart sounding set of grammatically arranged words often seen in research contexts.
Are you using gen AI to avoid doing deep work?
This is an important question. What are you really trying to achieve? What are you missing out on by perhaps depending on this tool too much? What alternatives are there? Might a healthy session of exercise followed by some deep reflection offer you more choices to follow up on?
For literature reviews, try gen AI as an additional tool to your own research
This is a similar point to the first one (and maybe the third one). We know how gen AI works, reporting mostly on the more obvious points and not always accurately. This is why it might be helpful as a means of checking on what you might have missed in your lit review. You will still need to check that what the tool reports to you is actually the truth.
Outsource some tasks to AI
There are plenty of forums where this is being discussed specifically in relation to your field of interest so as people reading this have a diverse set of research passions, I will not mention them here.
Check the target journal guidelines
Review the guide for authors about whether the journal editors accept gen AI use in the writing process. This is changing fast so it is well worth keeping up to date.
Make sure you know if the tools you use save your information
Ownership of the generated output is probably one of the most contentious aspects of gen AI. These tools tend to have stated terms and conditions, some of which are clear and others are ambiguous. It is well worth it to ask colleagues if you are not sure if the AI proprietors claim ownership of the output. Confidentiality and security are highly compromised when using some gen AI tools. Your information could be sold to others. This is probably as important as the first point.
Gen AI is changing fast and my own ideas about using it have changed over time as well. They will probably continue to evolve as gen AI becomes capable of more and more.