Speaking of AI, I wish the rest of the University of Chicago would adhere to our Law School’s new dictum, reported in the article below from WGN9 (click to read):
An excerpt:
Some students at the University of Chicago will have to learn how to do their work the old fashioned way, without their phones or a laptop, as the school is rolling out a new policy to curb the use of AI.
The students starting law school this fall are the first class who had access to AI platforms throughout their whole undergraduate careers, so administrators are making changes to the curriculum to meet the moment.
Cell phones, laptops and tablets will be banned from certain first-year core classes, such as Constitutional and Criminal Law, starting this fall.
“[The students] will be in a device-free classroom,” University of Chicago Law School Dean Adam Chilton said. “We need to be training our students to make sure that they can think critically and rigorously and independently without relying on AI.”
However, Chilton said instructors will layer AI into other courses because it’s a tool students will use in their careers and they need to learn how to incorporate it effectively and ethically.
“So that’s the challenge we face, both making sure that they’re not just solely relying on AI as a crutch and not learning to think for themselves, but also not just shutting it away, pretending it doesn’t exist and sticking our head in the sand and ignoring the technology,” Chilton said.
Yes, of course, you can use AI to look up facts, but given that students can also use it to write, I think it’s time to ban take-home assignments as well, for it’s pretty much impossible to detect stuff written by AI. How do they learn to think for themselves? Studying and pondering at home, but demonstrating their abilities in class, more or less via the method Professor Kingsfield used in “The Paper Chase.”
Now of course this isn’t all that realistic, but I do like it.
I was asked, after I sent this to a colleague, whether our Law School requires the LSAT standardized test. The answer is that it requires one of three tests:
Applicants have the option of submitting the Law School Admission Test(LSAT), Graduate Record Examination (GRE), or Graduate Management Admission Test (GMAT) (if applicable – see below for details) as part of the Law School’s pilot program.
We are accepting the LSAT-Flex, GRE General Test at Home, and GMAT Online Exam to satisfy the standardized test requirement. For the GMAT Online Exam, please see the details below.
You must have a current LSAT, GRE, or GMAT score (earned within the last five years) on file before the Admissions Committee will evaluate your application.
This is good, and should also be followed by the University in general. We have a “test optional” policy that operates according to the bizarre “no harm” guideline:
There is no minimum GPA or required test score. At UChicago, the admissions committee considers a candidate’s entire application—academic and extracurricular records, essays, letters of recommendation, and optional testing according to our no harm policy—and there’s no one piece of information that alone determines whether you would be a good fit for the College. You can learn more about this contextual review process here.
“Holistic” admissions are a euphemism for “no standardized metric”, but of course, given the uselessness of grade-point-averages as well as letters of recommendation, holistic admissions is nearly a completely subjective process.
The “no harm” policy apparently means that even if you submit a lousy standardized test score, like a bad SAT, it will not hurt your admission, nor will your failure to submit scores. To me that sounds bogus unless the school pays no attention whatever to standardized test scores. If it does, then a high SAT score will help your admission, which means that a high scorer has an advantage over a low scorer. And if that’s the case, failure to submit scores will also give you a relative disadvantage (such a failure also usually means that the candidates doesn’t think his/her scores would help them).
There is no rational argument I can see for a failure to submit scores, as they have now become more valuable than high-school grades in assessing future potential (grade inflation has eliminated the value of grade-point averages). There is only one reason why the University of Chicago, along with many other schools, deep-sixed the use of standardized tests, and we all know what that is.
Fortunately, the savvier schools are either re-adopting mandatory standardized tests or thinking about doing so. For what other metric can be used to judge all applicants against each other?
I loved The Paper Chase and Professor Kingsfield, particularly when his humanity occasionally broke through.
Having a device-free classroom is a good idea, but will it last? And it’s only a piece of the puzzle, as student’s can still use “knowledge aids” outside of class. We’re going to have to figure out how to incorporate AI into higher education, not remove it. That’s the key challenge. It’s truly a disruptive technology, with all that entails. And it’s going to pervade everything.
It’s a little rough on students these days to ban laptops, since it seems a lot of them use them for note-taking. They should make those classrooms Faraday cages. Orientation at the Law School (and the College) should also include a piece on A.I. and the importance of confirming its output before using it.
I was going to mention the note-taking. I don’t think my 23 year old son has taken hand notes for years. Starts law in the fall … writing notes will be a shock!
In my field, I could not really ban computers from the classroom, because of note-taking. But I could ban them in the exam, and ask highly specific essay questions on how the assigned readings (journal articles) support and contradict each other and the material they learned in class, and which experiments might resolve any contradictions that they might find.
If the students can answer those questions, even if their prepping for the exam includes use of AI, I could be satisfied that their knowledge and understanding fulfills the requirements of the course.
But that is the advantage of teaching upper-level and graduate science courses.
Always impressive how dramatically students get their knickers in a twist when an exam requires short answers without reference aid.
You say ‘even if their prepping for the exam includes use of AI’, as if that’s not desirable. I think this is perspective is misguided because AI can act as an amazing learning accelerator. I often use it to refine my knowledge on a subject by opening a dialogue with it. I have a perfect example from the other day. I was talking to an ER doctor friend about a technical topic (my academic background is in pharmacology and biochemistry). My pal was talking about a consultant who refused to use beta blockers (to reduce heart rate and BP) in the treatment plan for an acute cocaine toxicity patient, because it could lead to unopposed alpha activity leading to hypertensive crises and death (Beta 2 receptor activation relaxes the smooth muscles of blood vessels while alpha activity constricts them). I said I thought it was a daft decision based on dogma and minimal evidence, and could put the patient at risk. I remembered reading a brief case report years ago but nothing to back it up. I also pointed out that selective B1 blockers (which can produced the desired effect without reducing smooth muscle relaxation) have existed for years. Making it even dafter.
I went home and spoke to ChatGPT about it, asking highly technical and probing questions and getting superb evidence backed answers. It was like talking to an expert for 30 minutes and coming out with a much clearer picture of the issue. It would have taken me many days to get the same result by other means. FYI, it is largely dogma, yet it remains standard teaching on the subject in many medical schools.
I know that was a tangent, but it’s neccessary to show how good and helpful LLMs can be on difficult technical subjects. I use it all the time for this sort of thing and i have learned enormously while doing so.
In specialized fields I disagree. Yes your example is helpful but were there any caveats like, “was helpful in 75 patients, had no significant effect in 10, and caused severe hypertension in 8” because if it didn’t, your ChatGPT failed.
There has never been any treatment 100% effect and without negative outcomes. Those studies do not exist outside of China.
There has to be a downside or I would not believe anything stated by anyone or anything.
And I will refrain from discussing the reasons not to use increasing doses of sedatives as that is a different discussion.
PS.
There is now an AI hallucination database for lawyers who present a court with AI derived, nonexistent citations. Just short of 1600 at last count. The courts have not been amused.
Buyer beware is all I can say.
https://www.damiencharlotin.com/hallucinations/
Take-home assignments are a major problem. On the one hand, some projects cannot be completed in the classroom, so there is no way to avoid them. On the other hand, some students, perhaps most, will cheat on them. I cannot think of a satisfactory solution.
Firstly, sorry for the length of this comment. I share the concern about what AI means for academic integrity, but I think we also need to be realistic about the technology. Generative AI is not going away, and there is currently no reliable way to prove that a piece of writing was produced by AI. The various “AI detectors” have unacceptably high false-positive and false-negative rates, making them unsuitable as evidence for something as consequential as accusing a student of academic misconduct. More fundamentally, the problem is epistemological rather than technological: once all you have is the finished piece of text, the process that produced it has been lost. Many different authoring processes can produce essentially indistinguishable final documents, so the finished artefact simply does not contain enough information to uniquely reconstruct how it was created. Any claim that “this was written by AI” can therefore only ever be probabilistic (with a huge margin of error), not demonstrative.
However, this doesn’t mean we should ban take home assessments. In particular
I don’t think abandoning term papers would be a good solution. In my own experience, and from watching my daughters at university, long-form essays and research projects are where the deepest learning often takes place. They force students to engage with primary sources, evaluate evidence, construct arguments, and learn how to conduct rigorous academic research. It would be a great loss if universities abandoned that because of AI.
Instead of trying to prove AI use—which cannot be done reliably—we should rethink assessment. One possibility occurred to me while reading this discussion. Keep the term paper, but pair it with an in-person assessment on the same topic. The difference between the two should not be treated as evidence of cheating, because there are many legitimate reasons why students perform differently in different formats. However, a large discrepancy could be used as an indicator that further assessment is appropriate. This is a also a great way to reinforce learning.
Ironically AI could actually be of immense help here. It could generate individualized question sets for each student based on their paper, which the lecturer or teaching assistant would review, make a selection and approve for the exam. Students whose coursework and examination performance differed markedly could then be invited to a short viva or oral discussion. The purpose would not necessarily be to catch them out, but to establish whether the submitted work genuinely reflects their understanding.
That also mirrors professional life. Engineers, scientists, doctors, architects and many others are routinely expected not only to produce work, but to explain it, defend it, justify their assumptions and respond to criticism. Those are valuable skills in their own right, quite apart from any role they might play in maintaining academic integrity.
So I don’t think the challenge is how to ban AI or detect it after the fact. The challenge is how to redesign assessment so that it measures understanding in a world where AI exists. I suspect that will be difficult, but if I can think of this potential part solution while reading this article I suspect superstar professors in academia are much more capable of doing so.