One University’s AI Guide for Students, Written with the Help of AI


“When permitted, you can use AI tools for help but not to do the work for you.”

That line appears in the recently released “AI-U“, billed as “a student guide to navigating college in the artificial intelligence era.” What would a typical student make of it?

Fortunately, the guide, put out by Elon University and the American Association of Colleges and Universities, contains some more definitive advice.

Two pages from “AI-U” from Elon University and the American Association of Colleges and Universities

Of particular use, I thought, was the following checklist:

Before you start

1. My school and my professor allow the use of the AI tools I’m considering for this assignment
2.
I clearly understand when and how I can use AI for this assignment

Doing the work

3. I am using my own thoughts, words and tone of voice
4.
I have checked sources generated by AI and properly cited any facts, statistics or quotes
5. I have critically analyzed the AI output and identified any false, biased or harmful information
6. I have documented where and how I used AI and cited that use according to my professor’s expectations
7. I have not used confidential, protected or copyrighted information

When the assignment is complete

8. I can explain my findings and demonstrate full understanding without the aid of AI
9. I can prove what sources I used and how I verified the information

Thoughts on this checklist are welcome. A question that arises, of course, is how to encourage students to comply with it.

This term I am putting a fair degree of emphasis on #8. Students in one of my lower-level courses work on a multi-part research project throughout the term, and the penultimate part is a graded one-on-one meeting in which the student gives a brief presentation and then is subject to extended questioning from me about it. They’ll have to know what they’re talking about, and so it won’t be good enough to have memorized an AI-written presentation. With 40 students in this course, these meetings are a substantial time commitment, and I don’t think I’d do this with a larger course without TAs. But I’m curious how it will go.

As for #6 (and possibly #3, though it is impossible to tell), I wish the authors of this guide had set a better example. At the bottom of the third-to-last page of the 21-page guide, they include the following:

Use of Generative AI in production of this guide:
Text outputs from ChatGPT, OpenAI; Gemini, Google; Claude, Anthropic; Perplexity, Perplexity.com; Meta AI, Meta. See www.studentguidetoAI.org/about for prompt details. Graphics and images from Adobe Firefly, Meta AI, Adobe Stock, Flaticon.com

At that webpage they declare the following text prompts they used in making the guide:

“What are some core principles for students regarding the use of AI at colleges and universities?” prompt. ChatGPT, OpenAI; Gemini, Google; Claude, Anthropic; Perplexity; Perplexity.com; Meta AI, Meta; 29 May 2024.

“Based on your knowledge of the AI revolution and the prospects for transformational change in employment and daily life, please produce a complete guide for 2024’s college freshmen to help them navigate their higher education and prepare for a future in an AI-influenced future. The guide should include at least three central themes, explain the importance of each idea and include practical tips for students to follow in pursuit of the themes of the guide.” prompt. ChatGPT, OpenAI; Gemini, Google; Claude, Anthropic; Perplexity, Perplexity.com; Meta AI, Meta; 29 May 2024.

“Create a simple explanation to promote AI literacy among students, using the concepts of machine learning, deep learning, generative AI, large language models, LLM implementations and AI productivity tools.” prompt. ChatGPT, OpenAI; 27 June 2024.

“How are employers using AI tools in managing position searches and what are the implications and recommendations for students use AI to facilitate their searches for positions after graduation? Prompt. ChatGPT, OpenAI; Gemini, Google; Claude, Anthropic; 19 June 2024.

How is asking an LLM to, for example, “please produce a complete guide for 2024’s college freshmen” as one is writing such a guide an instance of how to “use AI tools for help but not to do the work for you”?

I suppose the authors of this guide did not explicitly say they would be following its applicable suggestions in its creation, but you’d hope they would, no?

Additionally, while the authors list the prompts they used, they do not state anywhere in the guide itself which parts of the text are based on (or are simply tweaks of, or bald reproductions of) the answers their tools gave to these prompts. So they cannot be said to have adequately “documented where and how [they] used AI and cited that use.” They could have provided the relevant details, and in doing so, modeled the behavior they’re recommending for students. Perhaps the next edition of the guide will correct for this.

What guidance are you or your university is offering students about the use of AI tools for their coursework? How is it going so far?

guest

19 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Snow
Snow
1 year ago

From what you’ve pointed out It seems like a continuation of “do as I say, not as I do” type of instruction model which does not inspire adherence.
As a recent graduate, I have come to understand that it depends entirely on the nature of the student. Is the student there to really learn or just get the degree to get a job. By learn I mean pour over the materials, grasp the concepts and able to have a significant and meaningful dialogue about it.

Higher education gives the essence of a “check the box” for some. So, will these instructions on how to use AI change that, probably not. If it helps a student—who is there to just check the box—do the work faster and without having to think, they will comply maybe as much as the publishers did in the creation of this guide.

As a possible solution, I wonder if having students declare (separate from citing) whether AI was used and then have a set of long form response questions about the extent of use, if any. This could be used to support and save time during your students presentation and questioning. My thought is that this would prompt reflection, discourage honest students, and give pause to less than honest students considering flagrant deception. (Just a half baked thought.)

Cheers and best wishes.

Ian
Ian
Reply to  Snow
1 year ago

“it depends entirely on the nature of the student. Is the student there to really learn or just get the degree to get a job.”

This seems exactly right to me. It appears to me that students who want to, or are willing to, learn are decreasing while students who are in college to check boxes in order to get a degree are growing, clearly a problem if true.

A former colleague referred to it as an unsettling mix of deep cynicism (about college) and utter naiveté (about adult life) in many students. Lack of good faith engagement from an increasing number of students in GE humanities courses, even if good faith only used to get you through week two or so, effectively guts the project of the lower division humanities curricula.

How we got here is a long and complex story involving large-scale economic issues dating back to the Reagan years, long-term political issues dating back to the Nixon era, austerity, missteps within fields, and administrative bloat.

Most—if not all—of these factors were and are more or less out of the hands of actual faculty in the humanities. But I’m struck by a line in David C.K. Curry’s recent piece in The Chronicle. Curry writes of the disaster at SUNY Potsdam:

While faculty members in programs awaiting “realignment” dealt with sleepless nights, many of my colleagues who escaped the cuts reacted by keeping their heads down. Department chairs, understandably, have tried to shelter their programs and staff from the coming storm, which, unfortunately, places programs in direct competition with one another for students. This bunker mentality plays right into the hands of the administration.

I want to talk about the tendency of faculty to keep their heads down, focus on their own departments, and enter a bunker mentality because it’s exactly what’s been happening since at least the 2008 financial crisis. Maybe this is just the way people are, or maybe it’s specific to profs, maybe it’s specific to humanities profs, but in my view there has been a complete failure to engage in any kind of collective work for disciplines with declining majors, funding, hiring, and so forth.

The only thing faculty could have (and I do think it’s too late, at least in the short term) would’ve been to put themselves at risk as the crises occurred by taking vocal and collective action. While I admire and agree with the sentiment of faculty being arrested for, i.e., protesting in favor of Palestinian liberation, I’m not convinced that the impact of doing so is going to change a lot.

I’m left to ask where this passion and exercise of political speech was when higher ed became a degree mill digesting smart, promising people and turning them into PTSD-affected adjuncts, or community college teachers with PhDs, or “alt-ac” refugees? Where was the organized resistance to accepting paltry little bones of completely exploited adjunct labor in order to support a shriveling department? Where were profession/field-wide initiatives to get grad students properly employed? Where was the realization that hiring new faculty based in large part on the elite status of their alma mater causes a recursive death spiral?

These things didn’t have to happen, or even if they did have to happen, they didn’t have to happen with a whimper rather with a fight. The problem as I see it is that for many faculty the “system” more or less works to provide them with a good job and a good life. What they don’t want to do is rock the boat, and as a result we get the bunker mentality, the head in the sand, the ultimately vain attempts to rescue one department out of half a dozen. But the thing is that the system flatly doesn’t work for anyone else in many cases.

You don’t have to be a Marxist to see that departments that make regular use of adjunct labor are supporting themselves on pretty execrable exploitation of others. That bargain should never have been made morally or even practically. Entire fields have given up any claim to moral rectitude (which matters when you’re in the humanities!), supposedly “critical” paper Marxists collect their salaries and awards while their former students grind for a pittance without health insurance or retirement, and nepotism and status reign supreme over all.

The academic humanities had multiple chances over the past two and a half decades to take up a fight for survival and for moral and intellectual standing. I can’t think of a single time they did so.

All sorts of external factors caused these problems, but cowed faculty and disciplines simply failed to address any of them substantively. It’s awfully difficult to maintain your moral standing when you fail to stand up for the most vulnerable, when you appear cravenly self-interested to those who see the writing on the wall, when you clearly aren’t willing to stand up for the principles you espouse.

Last edited 1 year ago by ikj
Kenny Easwaran
Reply to  Ian
1 year ago

I’m not convinced that students that want to learn are decreasing – just that students who are in college for some other purpose are increasing, and so students who want to learn are going to be a smaller fraction of our experience.

ian
ian
Reply to  Kenny Easwaran
1 year ago

This is probably true. Perhaps we could say there are fewer students who are willing enough to engage with learning outside of their majors with good faith due to a widespread cynicism based on secondary teaching to the test, a difficult job market, shortening attention spans, discourse about the value of higher ed, rising tuition costs, and so forth.

I don’t actually blame the students at all. It’s not a value judgement. But it can make GE teaching very disheartening, a fact that seems to be widespread.

Kenny Easwaran
1 year ago

I think some of the criticism you raise at the end about following their own advice is possibly a bit too harsh. What would it look like to say “which parts of the text are based on (or are simply tweaks of, or bald reproductions of) the answers their tools gave to these prompts.”

I think it comes down to the ambiguity of “I have documented where and how I used AI and cited that use according to my professor’s expectations”. They’ve certainly cited their use of AI, and documented something about how they used it. They didn’t exactly document *where* they used it, but I’m not sure how much of that makes sense. If their prompts produced several full-document-like objects, and several lists of issues that should be considered, and they made use of some of the organizational structure from one or more of the AI outputs, and included a point that they had forgotten initially (but had thought about months ago) because it showed up in two of the AI outputs, what would be a good thing for them to say?

I certainly wouldn’t want final documents to become more about the metadata and less about the document itself!

I think this is related to some ambiguous questions about citation in traditional bibliographies. I’ve heard some advice say that if I discovered a paper because of a reference in a second paper, and I end up citing the first paper in my work, then I should cite the second paper as well. I’ve never done that unless I also quote or discuss the second paper, because I think these additional citations make it harder for readers to get at the ideas that matter for the argument (even if they make it easier for future historians to discover the history of my ideas).

Kenny Easwaran
Reply to  Justin Weinberg
1 year ago

This seems right.

David Wallace
David Wallace
1 year ago

I suppose the authors of this guide did not explicitly say they would be following its applicable suggestions in its creation, but you’d hope they would, no?

Why?

The main reason to constrain AI use is because if students use AI to write their essays for them, it will interfere with their learning. But the reason the authors wrote this guide presumably wasn’t to improve their ability at writing guides for students; it was to produce an actually good guide for students. If in fact ChatGPT4 can just write a better guide than they can (or more realistically, if it can provide the skeleton of a good guide which they then tweak a bit) I don’t see any reason not to use it. At the very least, the issues are very different.

David Wallace
David Wallace
Reply to  Justin Weinberg
1 year ago

I’m not even sure about that. Organizational documents rarely display the details of how they are written, nor is there an obvious need for them to. (Pitt’s academic misconduct policy doesn’t, for instance – it says under whose authority is was issued, but not who actually wrote it.)

Antonina Shachar
Antonina Shachar
Reply to  David Wallace
1 year ago

But is it “an actually good guide for students”? How would they evaluate it if they cannot compose one themselves? And if they can compose one themselves, why bother with AI?

PBB
PBB
1 year ago

Point 6 (“I have documented where and how I used AI and cited that use according to my professor’s expectations”) raises the question of what citation expectations we should be setting for our students. Lately, I’ve been thinking of adding the following to my future course syllabi:

“Policies on artificial intelligence (AI) use: 

Presenting AI-generated text as your own work is a form of academic dishonesty and is always prohibited. 

If you use a generative AI tool (e.g., ChatGPT, Gemini) to assist you in any way in completing an assignment (including online discussion posts) for this course, you must do two things: (1) cite the AI tool in your assignment as thoroughly as you would cite a traditional academic source, and (2) submit a separate document with screenshots displaying all of your interactions with the AI tool that are relevant to the assignment (e.g., if you have a conversation with ChatGPT to assist you with the assignment, you must submit screenshots showing the entire conversation). If this information shows that the assignment you submitted does not sufficiently reflect your own work, this will lower your grade (just as overreliance on a traditional academic source, at the expense of doing your own work, would lower your grade). If there is evidence that you have used a generative AI tool to assist you in completing an assignment but you do not sufficiently cite it and submit the relevant screenshots, I will treat the case as a suspected case of academic dishonesty.

Note that generative AI tools often give vague, false, or misleading responses (e.g., making up academic sources that do not exist; claiming an author said something that they never said). It is your responsibility to check your work for accuracy. If you submit work that includes mistakes due to AI use, those mistakes may impact your grade.”

I read Justin’s response to Kenny’s comment as being very much in the spirit of requirement #1 (“cite the AI tool in your assignment as thoroughly as you would cite a traditional academic source”). The reason I’m thinking of including requirement #2 (“submit a separate document with screenshots displaying all of your interactions with the AI tool that are relevant to the assignment”) is that I want AI use to be subject to the same standard of transparency as the use of traditional sources. When a student cites a published article in their paper, I can go look at the article in order to check the student’s paper for accuracy and honesty and to assess how much their paper reflects their own work; I want to be able to do the same thing with any AI output a student uses as a source. And I should note that the tone of most of the rest of my syllabus is a lot friendlier than what I copied above; basically, I try to convey that I care about my students and am committed to helping them succeed (and supporting them more generally if needed) but also that I’m a hardass about a couple things, including academic dishonesty. But perhaps the tone of the AI policies is still overly harsh?

What do people think of these syllabus policies – is there stuff I’m getting wrong, or stuff I’m missing?

Justin, thanks very much for this post and discussion!

Justin Fisher
Justin Fisher
Reply to  PBB
1 year ago

I think this policy wouldn’t be practical for many “ideal” use cases of generative AI, e.g. where the student engages in Socratic discussion of a topic with AI to explore it, or where a student engages in many iterations trying to craft a prompts that get the AI to produce useful results, copying and pasting out useful tidbits as they appear.

You presumably don’t make students turn in a list of all the mistyped words that their spell-checkers proposed new spellings for, nor all the quasi-grammatical expressions that their grammar checker suggested new phrasings for, nor all the prompts they typed into the their calculators along the way. I don’t think there’s any clear principled line between this and many legit uses of generative AI in exploring a topic and figuring out what you want to affirm about it.

praymont
praymont
1 year ago

Here’s an excerpt from the American Psychological Association’s policy on citing AI:

“If you’ve used ChatGPT or other AI tools in your research, describe how you used the tool in your Method section or in a comparable section of your paper. For literature reviews or other types of essays or response or reaction papers, you might describe how you used the tool in your introduction. In your text, provide the prompt you used and then any portion of the relevant text that was generated in response.

Unfortunately, the results of a ChatGPT ‘chat’ are not retrievable by other readers, and although nonretrievable data or quotations in APA Style papers are usually cited as personal communications, with ChatGPT-generated text there is no person communicating. Quoting ChatGPT’s text from a chat session is therefore more like sharing an algorithm’s output; thus, credit the author of the algorithm with a reference list entry and the corresponding in-text citation.” (Timothy McAdoo, “How to cite ChatGPT,” Feb. 23, 2024) https://apastyle.apa.org/blog/how-to-cite-chatgpt

Meme
Meme
1 year ago

If you’ll permit me to be somewhat un-philosophical about this—meaning I just want to vent without offering an argument—does anyone else sometimes feel that we shouldn’t cede any academic ground to GPT and other LLMs? That there is something in-principle profoundly messed up about corporately trained writing tools being even occasionally or lightly used by students? That the whole thing feels almost (hyperbole incoming) satanic, thought-stopping, destructive to what makes us human—the spontaneous, creative, critical expression of ideas? I know that AI is probably inevitable, I understand the arguments that we should embrace (while mitigating the effects of) that inevitability, I recognize that I’m being a starry-eyed romantic, and so on. But sometimes it feels like this is the sort of thing that we, especially as philosophers, should be resisting irrespective of the (possible) futility of doing so. I don’t man, shit’s a bummer.

Last edited 1 year ago by Meme
Meme
Meme
Reply to  Meme
1 year ago

Oops, should say “I don’t *know* man” at the end there. Guess I should have used GPT after all.

Steve K
Steve K
Reply to  Meme
1 year ago

I can totally understand this feeling, if you want to get more philosophical about it you might enjoy Shannon Vallor’s “The AI Mirror: How to Reclaim Our Humanity in an Age of Machine Thinking” (2024). I came across it in this discussion: https://www.abc.net.au/listen/programs/philosopherszone/shannon-vallor-what-is-ai-doing-to-our-humanity-/104068090

Antonina Shachar
Antonina Shachar
1 year ago

One problem I’m seeing is the conflation of AI with generative AI.This whole guide seems to be referring to generative AI, and should say as much. #4 shouldn’t even appear – it encourages one to generate sources, statistics, and quotes as long as they are “checked”, which seems like an insane waste of time when you can go to a reliable source directly and skip the need for verification.

I encourage my students to use generative AI to tease out concepts and develop their ideas (like a conversation partner) – but to avoid relying on it for generating facts, stats, quotes, or sources. There are much better tools out there for the latter. One fantastic use case I found for it is to generate analogies relevant to the student’s own life, to make the content more memorable and comprehensible. In this case the student can evaluate whether the analogy makes sense or not.

Kenny Easwaran
Reply to  Antonina Shachar
1 year ago

It can be useful in generating sources, just as asking a colleague for sources can be useful even if they sometimes get the title wrong or misremember which paper they meant to cite – you do have to go find the source and check it yourself, but finding the source is hard if you haven’t already read it yourself, and it doesn’t specifically use the same word so that Google can find it, and a human or a language model can often help with that sort of thing (despite their different flaws).