Should University students be discouraged from using generative AI to write their essays?

By Dr Julian Koplin, Lecturer in Bioethics at Monash University, and Dr Craig Stanbury, Postdoctoral Researcher at the Ethox Centre, University of Oxford.

Introduction

Below we have written an ‘open letter’ to university students who use AI to write their essays. While the letter reads as addressed to one student, there is no specific student to whom we are addressing it. The student in this letter is ‘abstract’, and the intended addressee is every student who relies on AI to do their work. 

The purpose of this letter is to show sympathy for students who feel they need to use generative AI to get a degree, particularly at a time where refraining from doing so might disadvantage them relative to other students. At the same time, we want to challenge students to reflect on their values, and to consider whether outsourcing their learning to AI reflects those values.

We have been teaching at various Universities for over 10 years and are deeply concerned about the situation our students find themselves in. We are also concerned with how typical responses to this situation criticise students for cheating. We think that more attention needs to be paid to the incentives students have to cheat with AI, and the costs to students themselves for using AI – often in ways that their university encourages.

At the end of the letter, we have listed reasons for and against using AI to write essays. We would love to hear your thoughts. Where do you stand on the matter? Should University students be encouraged or discouraged from using generative AI to write their essays?

To the Student Who Used ChatGPT to Write Their Essay

Hi,

We recently graded your work. We say “your work”, but we’re not sure if that description fits. There was, as far as we could tell, very little of you in it. There was nothing to suggest the presence of a living, breathing writer who cares about the world they have to live in. We tried to discern signs of life in your essay, hoping to see a trace of you peeking through the curtains. Alas: there was nothing.

The essay used so many words and phrases associated with AI chatbots.^[1] The problem you were writing about was “intricate”; the solution, “quietly profound.” You surveyed many sources, and they each had some sort of vague point that you endorsed on your way to a vapid conclusion (essentially, that the issue is important, and people should keep thinking about it.) The essay was competent, but entirely unremarkable. We did what any ambivalent marker would do in such a situation and awarded it a Credit.

You aren’t in trouble. We’re not allowed to use AI detection tools, and in any case, they don’t work.^[2] Perhaps you also thought AI was allowed in this subject. It would be an understandable mistake. Many subjects do allow the use of AI, and our institution even provides you with free access to certain AI chatbots. It encourages the use of this new technology because after your studies you will probably enter a workplace where the use of AI tools is accepted, and maybe encouraged. It is good, the university thinks, for you to build experience with these tools. By turning your assignments over to AI, you are merely following the path that society has set for you, with our institution cheering you along.

But, there are costs.

Emerging evidence suggests that relying on generative AI impedes various forms of learning[3a][3b][3c][3d] and critical thinking.^[4] Some scientists speculate that heavy use could even impair certain cognitive capacities.^[5] In delegating your writing to ChatGPT, you are trading genuine learning for a gentler workload.

And this reduced workload may be a mixed blessing. Cognitively effortful activities support human wellbeing:^[6] It is rewarding to complete difficult tasks. By contrast, intensive AI use can produce what one group of researchers has labelled “AI brain fry”: a foggy, buzzy mental fatigue reported by workers who spend their days overseeing and interacting with AI systems rather than attempting tasks themselves.^[7] We worry about where the habit of delegating difficult tasks to AI might leave you.

We also worry that this habit might leave you less yourself. We ask you to write essays not because we have any particular desire to read 200 students’ opinions on whether there is a moral obligation to donate more money to charity, but because writing is a wonderful tool for developing your thinking. As Joan Didion put it in 1976, “I write entirely to find out what I’m thinking, what I’m looking at, what I see and what it means.”^[8] 

Later research has shown that she is right: writing does indeed help with these tasks.^[9] Doing the writing yourself – trying to explain in full why you believe something – can help you to assess just how good your reasons are for holding that belief. It helps increase your confidence in an idea, and helps determine whether it holds up to scrutiny. Neither of us are arrogant enough to think we have Didion’s way with words, but both of us have benefited from expressing, testing, and revising our ideas throughout the writing process. 

ChatGPT probably cannot help with this task. More than that, ChatGPT can get in the way of finding out what you think. One apparently “responsible” use of ChatGPT is to enlist it to convert dot points and rough paragraphs into polished prose. But, AI chatbots tend to be highly agreeable, even sycophantic[10]: they can put a convincing sheen on bad ideas.[11] It is also easy to read their output and think, yes, this is what I was trying to communicate all along. 

However, ChatGPT (despite its occasional protestations to the contrary) is not neutral.[12] In producing polished text, it supplies assumptions, emphases, and a moral vocabulary shaded by political leanings embedded in the model.^[13] These values are set by the companies that build these systems. Even at the best of times, the politics of a company like OpenAI might not match your own. But these are not the best of times: AI companies now face pressure from political actors, including the Trump administration, to reshape these models in line with their own preferences.[14] Even those using ChatGPT “responsibly” – for brainstorming, copy-editing, and feedback – may come to mistake the tool’s way of thinking for their own, and so inherit someone else’s moral commitments.

Indeed, there is evidence that AI chatbots can shape how people think. One group of researchers discovered that even seemingly neutral information provided by these tools often sways users’ political opinions.^[15] Another, that the very language they use is shaping how humans speak.^[16] More worryingly still, there is evidence that when students use them to help draft essays, it results in a kind of homogenisation of their writing.^[17] Given the relationship between writing and thinking, it would be unsurprising to find that ChatGPT homogenises your thinking, too. 

This, we believe, is the biggest cost. You are a thinking, dreaming, living human. Your thoughts and your inner world are unique. ChatGPT has been described as a “blurry jpeg of the web”:^[18] we want something better for your inner life than this. 

You recently graduated high school. Free from the stewardship of your teachers and parents, you have now joined the teeming mass of adults trying to work out how to live. This involves curating a life that is suitable to the particular person that you are. The more that ChatGPT shapes your thinking, the more it erodes the edges of who you are. The uniqueness of your way of thinking, what you love and are passionate about, and the things you believe in – all these things are flattened into everyone else’s view. 

It breaks our hearts (and resolve as educators) that the educational system you find yourself in encourages this AI-assisted flattening. We doubt you are happy about it either. When we speak to students, many are openly cynical about generative AI; some are downright despairing. Still: whatever the drawbacks of AI usage, you are undoubtedly very busy. You are juggling many responsibilities: work, study, family, and relationships. And generative AI tools are convenient. Writing an essay is hard, time-consuming work. It is difficult to pull off in the way that will get you good grades. AI-generated output often reads well – certainly, your AI-produced essay is more polished than anything we produced in our first years of university study. If you want good grades, AI may well be the way to go.

And perhaps you feel lost at university. Classes are big; attendance is optional. We would like to offer you support with your writing – not that we are gurus, just fellow travellers with a few more years experience – but one of us is a casual academic on a contract that doesn’t include any student consultation hours; the other, a lecturer tasked with overseeing hundreds of students, among other demands. We don't recall seeing you in class; maybe we've never even met you. With nobody looking out for you, it is no doubt reassuring that you can turn to ChatGPT for help. 

For all this, we are sorry. We are sorry that outsourcing your thoughts to ChatGPT seems like a reasonable thing to do, and we are sorry that it would be asking too much of you if we told you not to. We are sorry that the system has let you down, and that we cannot help you more.    

Nevertheless, we also wanted to encourage you to keep thinking for yourself. There is probably some part of you which does care about the things you are studying, and about your role in the world. You very likely care deeply about moral problems and how to personally navigate them. You were very likely inquisitive as a child, as indeed most children are: we imagine you as someone who read, questioned, doubted, made mistakes, and experimented with different ways of thinking. You must have encountered ideas that excited you, or grappled with the human condition, and enjoyed using your imagination to solve problems. If this was (or is) you, there must be something about education and learning that you value. 

Whatever the case, you certainly have thoughts about our world. These thoughts are worth developing. You get to take part in the long conversation humanity has with itself about how to live and who to be. We do not want to see ChatGPT take your place in that conversation. 

As society tries to mould you into a certain type of person, and as AI tools start exerting their own influence on your thinking, remember to keep interrogating every idea you receive. Even though the system is stacked against you, please don’t allow yourself to be flattened into a hazy, low-resolution version of what an AI system thinks a human should be. 

So, before we sign off and grade the next essay, we wanted to ask a few questions that only you (and not an AI tool) can answer. We ask these questions in full knowledge that life is often very hard and very busy, and that much can get in the way of sitting down and thinking about how you want your life to go. Nonetheless, we suggest that you turn off your phone, go for a walk, sit in a park, and think: 

●        What are your values? What do you think is worth striving for?

●        How do you want to live? How might this differ from the default template of a tech-saturated, 21st-Century life?

●        When, where, and to what extent will you let AI tools do your thinking for you? Can you stop them from dominating your thoughts? If they do change how you think, is this acceptable?   

Wherever you go in life, whatever job you end up getting, and however effective AI technology becomes, the answers to these questions should only ever be yours - not someone or something else’s. 

Sincerely,

Your Concerned Tutors,

Dr Julian Koplin and Dr Craig Stanbury.