Imagine you’re struggling to write your thesis and wake up one morning to discover there’s an artificial intelligence (AI) tool that can weave your less-than-perfectly crafted prose into scholarly gold. Would, and should, you use it?
I recently received an email from a PhD candidate who’s wrestling with this question. With her submission deadline fast approaching, she asked my advice on how to approach using AI-powered writing tools to responsibly refine her thesis without committing academic integrity violations. Specifically, she wanted to know if AI tools with grammar and rephrasing capabilities would be considered plagiarism or an act of academic misconduct.
Those of you who’ve been following my thoughts on AI tools will know that I’m a big fan of leveraging AI to help augment our abilities. But I also stress in all my AI training sessions that there’s a line – albeit sometimes a fuzzy one – between using AI to enhance your contribution to knowledge and using AI to produce outputs that have little in common with your domain knowledge or level of expertise.
As more graduate students turn to AI tools to augment their writing process, I thought it timely to reflect on the topic in this blog post. Noting that what follows constitutes my opinion and not formal advice.
When considering the use of AI for the creation of scholarly outputs, always seek guidance from your university!
Could versus Should
First, let’s address the question of whether you can generate significant amounts of text with AI tools and paste them into your thesis without detection.
Yes, you can.
But should you?
I think not.
Asking an AI assistant to write entire sections of text and then pulling a sneaky Ctrl+C and Ctrl+V is problematic from several standpoints.
While it may seem expedient at first glance, this approach robs you of deeply engaging with the conceptual connections across your research and limits the knowledge growth that comes through grappling to articulate such links yourself. Allowing algorithms to do the heavy lifting might also preclude you from catching deeper insights that often emerge in the struggle to translate fuzzy mental models into clear text through ‘writing to think’.
Relying solely on AI to outsource your thesis writing rids you of the opportunity to build the foundational knowledge that will be required to effectively participate in academic discourse in your field in the future.
Furthermore, and perhaps the most cited reason for avoiding cut-and-paste output creation is the fact that claiming authorship over content you didn’t substantively contribute to generating (beyond writing an initial prompt to get an AI tool to generate the text) undermines scholarly rigour and integrity.
While intelligent algorithms can creatively recombine impressive vocabulary, they cannot replicate the context-driven sensemaking cultivated through your long journey with the literature and analysis.
Excessively depending on synthetic support is no substitute for rigorous scholarly thinking and writing.
AI Copywriting Versus AI Copyediting
You can think of the cut-and-paste scenario as an example of AI copywriting, a term I’m using to refer to the process of outsourcing scholarly writing and thinking to algorithms to generate text with minimal or no human intervention.
AI copywriting raises ethical alarm bells.
An alternative is AI copyediting.
AI copyediting refers to using AI tools to work with text you’ve already written to augment your scholarly writing abilities; using AI to gain feedback on your writing and suggest improvements.
In this scenario, you are working with text you’ve already conceptualised, and then improving upon the hard work you’ve put in to lift the written style, tone, sentence structure etc. to a higher level.
When done well, using AI as a writing assistant can significantly lift the standard of scholarly writing.
As an example, sections of text can be provided to an AI tool like ChatGPT alongside an instructional command that requests the tool to make suggestions for how the text could be improved. Upon receiving the output, you can then work with the suggestions to improve your text. In this way, you are using AI with the level of agency to adopt or ignore its suggestions and enhance understanding of how text could be improved which can feed forward into future writing efforts.
I believe that AI copyediting is an efficient way to employ a sophisticated editing assistant without the financial price tag associated with hiring a human editor. Thus, using AI for copyediting also levels some of the privilege advantages that wealthier students (whether that be due to their university’s resourcing or private funds) have concerning hiring external assistance in the form of human editors. Hiring a human editor can cost thousands of dollars, whereas using AI tools for editing can be free, at least from the perspective of an individual’s outlay of financial capital. See Kate Crawford’s book Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence for a deep dive into the wider costs associated with AI.
Appropriate Usages for AI Thesis Editing Tools
Concerning the email enquiry I received, I suggest applying editing features to streamline awkward phrasing, vary sentence structure, transition between ideas, adopt appropriate academic language and polish written work. I feel this is a legitimate use of AI tools that can move writing towards thesis-level quality while retaining your core ideas which is comparable to traditional human thesis editing and advising.
I also believe that leveraging the functions afforded by AI tools like summarising arguments to improve flow and transitions between chapters, catching minor grammar and punctuation issues, or paraphrasing very small sections of text for clarity represent responsible applications of AI.
Verify Your Institution’s AI Policies
As policies around AI usage continue to develop in response to rapid change, students wishing to use AI tools should keep an eye on their university’s guidelines. Some institutions already allow editing assistance when it’s accompanied by an acknowledgement. Please check your university’s guidelines!
Addressing Concerns Around AI Text Detection
Regarding fears over AI detectors concerning scholarly writing, I have a sustained scepticism that there will ever be a reliable or valid AI detection tool. For an interesting (and rather humorous) look at this issue, look at one Dr Lyndon Walker’s experiments with AI detection tools.
The AI Sniff Test
Many of us are operating in environments with an absence of clear guidelines on AI usage. I don’t blame universities for this situation. We are in a period of transition amidst the arrival of new technology and there are simply too many edge cases; academic integrity policies can never explicitly outline every single scenario in which AI might be used.
One way for students to navigate this period and make judgements about whether they should use AI to perform specific functional tasks that will contribute to the thesis document creation – whether that be generating entire paragraphs of text or using it to identify and fix typos – is the good old fashioned ‘sniff test’.
If something about using the tool’s output doesn’t seem right or makes you feel uneasy, listen to that inner voice – it often picks up on subtle signs of potential plagiarism or misconduct that should give you pause.
At the end of the day, you are submitting a thesis bearing your name. How you choose to use AI (if at all) to accomplish that task needs to be done in a way where you feel comfortable claiming authorship.
If in doubt, err on the side of caution.
If you ever feel yourself sliding into territory where it feels like AI tools are doing more work than you are, yet you’re claiming ownership of that work, it might be time to roll back your AI enthusiasm and save it to produce less career-critical outputs.
To receive more blog posts like this in your inbox, sign up for my email newsletter T3 to receive a dose of tech tools, AI tips, and training.