Thank you for your reflections. I am a junior faculty member at UC Irvine (I've just started this year!), and I generally agree with all your points on faculty comparative advantage, the stress on the peer review process, etc., but I wanted to share my reactions and in doing so, bring up two ideas for potential discussion:
(1) In the classroom: I am very sympathetic to returning to in-person exams over the take-home papers. Unfortunately, I feel that myself and my TAs are very stretched thin. I am teaching two undergraduate lectures of about 100 students each, and have one TA assigned to each, and who knows how resources will change in the coming years. My solution was essay-style in-class quizzes that were done on Canvas with tools to prevent use of other applications during class time—though I am not sure how well an "Einstein" could get around that—and we have proctored exams, for what it's worth. It is genuinely faster to grade electronic work compared to blue books (we have green books?) if my experience in graduate school is any indication. But I do wonder if there's anything inherently wrong with using AI for grading? For the record, I did not do it this term, but I do think there's an inherent asymmetry between students and the instructional team in that the former is asked to reproduce knowledge while the latter is evaluating it. I learn from my students, for sure, but universities as institutions seemed to be comfortable delegating a large portion of the grading to TAs, would this be too different? Put differently, did we in the past delegate grading work to TAs because the goal was for them to learn? Or to protect faculty research time? I guess my broader point is that this is likely to exacerbate existing inequalities across universities, and if UC Irvine, which obviously relatively well-resourced and I am in my own place of privilege...
(2) On AI-authored or even assisted research projects, you've pointed to lived experience in the AI University section, and it's not lost on me that you subtitled the piece "*field notes* from a tough year," but to me the research connection is there too. I am primarily a quantitative scholar that works with administrative data, and unfortunately, it does seem like in that space AI is in the chess + human realm (though I try and avoid the use of LLMs for writing per se and stick to using it for "engineering" tasks, as Anton Strezhnev eloquently discusses in his AI policy https://www.antonstrezhnev.com/ps813/syllabus.html)... but taking a step back, I feel that the context-building fieldwork I did in Colombia for my dissertation—again, I am not trying to pretend to be a qualitative or ethnographic scholar—was so immensely valuable and not something that AI can replicate. Sure, AI might be able to take fieldwork notes and churn out a paper, but not have that lived experience in and of itself. Similarly for the field experiment I am currently fielding with a co-author. Sure, AI can help with our code, but it did not build the relationships with the California Department of Tax and Fee Administration who is partnering with us on the RCT, much less implement the experiment itself, which the CDTFA is doing. It strikes me that work based on these lived experiences (whether qualitative or quantitative) will be more strongly valued, and for what it's worth based on my limited teaching experience, that seems to be what students really appreciate, too.
Julian thanks for these thoughts. A couple quick reactions-- on the research front, my hope is that the AI age will make the discipline put more value on qualitative work, interviews, fieldwork, etc. I agree that these are the things AI can't quite do and increasingly I find myself gravitating towards these methods. Re: AI grading, I take your point, and I wish universities would just hire more faculty so we could teach classes of small enough size to give the work the attention it needs. But point taken about inequities here-- I think there are probably some assignments (like MC tests, problem sets) where AI grading makes sense, but I haven't thought it through enough yet. Enjoy the first year of assistant professor life! Pressure aside, it's a nice chapter and it goes fast. - Rory
Thank you for your perspective, Professor. I am concerned about the degradation of learning with AI. It is and will damage students’ knowledge IMO. We need guardrails put up in all areas of AI. I think universities can and need to address it.
As far as demoralization, I point you to the recent book by Angus Fletcher “Primal Intelligence”. His research acknowledges the human brain is far superior bc of its creativity in story telling—much like how you describe your experiences in China and your college era. We can use our five (or 6) senses and we can teach a far superior visceral storyline.
His studies (w the US Army etc) alleviated my anxiety and renewed my hope in humanity vs computer.
What struck me most was not the future of AI, but the future of human relationships.
Throughout history, civilization has steadily reduced the cost of accessing knowledge—from books, to libraries, to the internet, and now to AI. Information and expertise are becoming increasingly abundant.
Yet when I think back on the people who shaped my life, they were rarely those who simply gave me information. They were teachers, mentors, and friends who offered perspective, encouragement, and human connection.
This made me wonder whether the true value of education was ever knowledge itself.
Perhaps as knowledge becomes abundant, humanity becomes the scarce resource.
And perhaps the future of education is not about competing with AI, but about rediscovering the human relationships that make learning meaningful.
Rory, as an emeritus professor who now only teaches in the classroom for a few weeks a year, I also have felt demoralized by the difficulty in figuring out how to evaluate student learning in the age of AI. I wish we could concentrate on teaching and advising and not worry about the evaluations — I wish everyone could take the course pass-fail instead of for a grade. This year my co-teacher and I are going to use more in-class quizzes, exam, and a group project to be presented in class instead of the take home essay assignments we once used both to evaluate students and to teach them how to organize and write a cogent argument. Replacing them with in- class exercises will drastically cut into the lecture and discussion time. I don’t see any way out of this dilemma, or the angst that I, you, and other academics now feel about our mission as teachers. Thanks so much for taking the time to write this thoughtful essay.
Susan thanks for reading and sharing your wisdom. There are many days I think we would all just be better off if we got rid of grades. Or just converted to something resembling Distinction, Pass, No Pass.
Your writing really moved me, professor. I think it's the time for us all to be worried about our education system. Though I'm not American, I can attest that the situation in China is quite similar. Primary and secondary schools here are a little less impacted since screen time is regulated and phones are still banned in a lot of the schools, but universities certainly face a far bigger crisis. The way you described a shell university gave me chills.
As I recall my college years in the States, I think I was the generation riding the tide. I lived through the stages where AI was a barely usable novelty to a genuinely useful research and editing tool as ChatGPT evolved. Yet, even as one of the heavier users of chatbots among my peers back then, AI remained a peripheral tool during our education. Today, the landscape is just completely transformed. It's deeply unsettling to imagine what higher education looks like now with students outsourcing their critical thinking to agentic AI. I know going into work probably requires you to be proficient in at least some of the AI tools, but I'm at least grateful for my own learning curve and the hardships I've experienced during uni.
Does our value still lies in our knowledge? In the stories we tell? Or maybe it's like what you said. Our value exists because we do. We need to hang on to the places we’ve walked, the emotions we've felt, and pass down to those who come after us in a face to face conversation.
Thank you, this article is both thoughtful and thought-provoking. It takes me back to my undergrad years. I began at UC Davis in the fall of 1968. Not only pre-AI, but pre-digital, though there were rumors of students wandering campus with large cardboard boxes carrying carefully stacked punch cards to the lab so they could print out the prized “smiley face.”
My poly-sci class had over 100 students. We sat in a large lecture hall, squinting at our professor way down at the bottom of the tiered seating. The large lectures were supplemented with much smaller discussion sections led by weary TA grad students who were somewhat more accessible. And ohhhh, those Blue Books. In hindsight, I wonder how many were ever read by the professor and how many grades were delegated to the AI of the day (TAs).
Years later, I found myself homeschooling my children variously, sometimes one, sometimes both. As I consider your comments, I reflect on my own motivations while supporting them through those years. The one thing that stands out for me is wanting to protect their curiosity. I wanted them to love learning. That still feels like a worthwhile guiding principle despite the treacherous waters of academia. Thank you for fighting the good battle.
I am in an interesting position in the private sector (investment management) where our company has been adopting AI at a rapid rate. Sort of reminds me of an "enhanced game" type vibes; use what's valuable to push performance to the limit. It's been interesting to see what skills are emerging as valuable when knowledge is readily at hand and complex, analytical tasks can be executed in an instant. Some of the main things, I'm finding, are (1) knowing what questions to ask and (2) knowing what to do once you have the answer. It's something of a buzzword, but "taste" is a distinguishing factor for both of these, which is often gained through years of experience and toil (i.e., many failed investigations gives you a sense of what is worth going after).
In terms of the new graduates, there are a few possibilities. One is that they are AI native and will mostly blow us away with their ability to use these tools. The other, though, is that they increasingly haven't done the difficult toil of working and failing that is required to develop taste on what is worth asking and what do with the answers once you have them.
Lastly, I love love Good Will Hunting, but maybe there are benefits to not having seen it. For example, this summer I would be excited about The Odyssey, but it's hard for me to get on board with an Odysseus from Southy.
Thanks Alex, I appreciate the perspective from the employer side. I think this generation will be more agentic than others and better at using the tools. The rest of us will still be stuck in our old ways. But as you’ve said, they might also lack some of the cognitive investment in learning different skills and ideas. I suspect the ability of people to write will degrade over time, as reading already has.
"Yeah, maybe, but at least I won't be unoriginal" Let that be the mantra of college students!
I think where humans still have the comparitive advantage is in synthetic reasoning, hyperpolation (as opposed to interpolation or extrapolation where LLMs excel), first-hand observation/data collection, and high-touch care. I think we need to think about how a university education can teach those skills.
I argue here (https://bobm858524.substack.com/p/the-work-of-humans) that knowledge discovery is an economic complement to knowledge transmission/diffusion. Since LLMs make knowledge diffusion cheap and easy, knpwledge discovery will be even more important in the age of AI. Universities, therefore, should lean into their research/scholarship missions, incorporate undergraduates in the process, and reunify their research and education missions.
This won't be easy, and you make the important point that academic journals are not set up to filter out the slop or handle the volume. Still. I think we need to think about the business model of higher education in the context of our education and research missions.
My wife is a full professor at a peer university of yours and I hear all of this nearly every day. It’s breaking her. It’s breaking her spirit and love of teaching. She spends more time investigating and pursuing honor code violations than students spend “writing” the slop that that eventually gets turned in. She has had to design an entire syllabus around the creation of hand made products in a writing class just to steer clear of the giant sinkhole in the road called Ai. She wants out but the golden handcuffs of a full professor job in academia are very very tight.
This is a very good essay. I'm a historian, but I understand your frustrations because I have discussed them extensively with my colleagues in Political Science and see the disciplinary differences. My take on this problem is here:
It follows your point that "professors need to focus on our comparative advantage over the LLMs." For me, that means getting in touch with my inner-English professor because LLM writing stinks, and it isn't going to get any better because it will never be written by humans.
I think it's worth underscoring that more proctoring could redress one of the major worries professors currently have -- namely that students aren't doing their own writing/thinking. So I'm definitely for that -- especially if that proctoring work could be done by testing centers/writing centers instead of burdening instructors with having to administer blue book exercises during class time.
One thing that deserves qualification in your essay is that what is worth celebrating in a Princeton education isn't the old style lecture (even if it's in Einstein's lecture hall). Instead, it's the precept system Woodrow Wilson inaugurated back in the early 1900s. That f2f interactive learning around a board room table isn't in danger of being hollowed out by AI. Princeton (and other HE institutions as much as they can) should keep doing that. It's the best way of keeping Silicon Valley's efforts to automate learning at bay.
Great post. I jokingly asked my son Eduardo (‘23) if he was your student. Right off the bat he said: “Yes! Rory. Chinese Politics. Took him Fall of 2022. He was great. Extremely popular with the students.” Loved that he called you “Rory”, not Professor Truex. So, seems to me you have got it right since before Claude, Rory. 🫡
Thanks for your reflections from a young-ish professor—many of the older one don’t get it.
Coming from a STEM background, I think we should evolve university to incorporate AI the way that engineering companies are doing so today.
I’ve always thought the point of getting an engineering degree is to get hired and have a career building cars, planes, rockets, etc. If we work backwards from that premise, it means that we need to learn the engineering principles and, just as importantly, the tools needed to build these products.
Engineering companies are using AI alongside existing engineering tools. By restricting students from using AI, we doom them in the “real world”.
Outside of STEM, I largely agree with your point. LLMs can replace much of the essay assignments if left unchecked. Likely going back to pencil and paper is the best short term approach as we figure out how to evolve university for the AI age.
Thanks for this reaction and for reading. I agree that this needs to be field specific and pretending its the 90s is not a viable long term solution, even in political science. They need to learn to use the tools. Next time around I plan on revamping my China course to see what they can build. @Andy Hall is very innovative on this front.
Rory Truex’s recent reflection, “Will AI Break the University?” (Truex 2026), offers unusually candid field notes from a professor confronting the collapse of the take-home essay. His piece reveals a deeper exhaustion: an educational model drifting toward mechanisation long before ChatGPT. AI did not destroy the university; it exposed how thoroughly industrial the modern university had already become.
Truex recognises that a large language model can now summarise the Chinese political system faster than any lecturer. What students remember is not content but personhood – his experiences in Beijing, uncertainty, humour, and invitation to lunch. The future university, he intuits, must rest on human relation rather than information scarcity. Yet institutionally he retreats to the blue-book exam: three hours, pencil and paper, closed book. The educational equivalent of reverting to horse-drawn carriages after the automobile.
The irony is striking. The intellectual traditions universities celebrate emerged in conditions impossible to replicate today. Ludwig Wittgenstein arrived at Cambridge in 1911 not through standardised credentialism but through force of curiosity and Gottlob Frege’s informal suggestion. Possessing only an engineering Diplom, he appeared unannounced at Bertrand Russell’s rooms in Trinity College. Russell, uncertain whether the young Austrian was “a man of genius or merely an eccentric,” nevertheless had the time and institutional flexibility to discover the answer through direct dialogue and a vacation essay. That older Cambridge operated in a small, elite ecosystem where faculty-to-student ratios permitted genuine mentorship (McGuinness 1988). (It is worth noting that this ecology was itself profoundly elitist; its flexibility was never designed for mass higher education.)
Such permeability has largely vanished. In the industrial university of 2026, professors manage massive administrative loads and lecture halls of one hundred or more PoliSci students.
Assessment is optimised for scalability and auditability. Handwritten examinations appeal not for pedagogical superiority but because they restore measurable control. The burden of proof shifts to the student’s body: cramped hands and stressed memory become proxies for authenticity.
This shift is especially punishing for neurodivergent students who excel in dialogue and iterative exploration. Would Wittgenstein himself thrive in today’s system? It is far from obvious. His development relied on prolonged discussion, obsessive inquiry, and personal mentorship.
Truex senses that the professor’s comparative advantage now lies in relational presence, yet remains tethered to institutional procedure. Tony Seba’s analysis of technological S-curves suggests that once a superior system gains velocity, the old order rarely survives incremental adaptation (Seba 2014). AI renders knowledge abundant. The university’s value can no longer rest on content delivery and credential verification.
The civilisational choice is clear: continue defensive proceduralism: surveillance, proctoring, timed exams, or rediscover education as dialogue, mentorship, and recognition of unusual minds.
Truex’s focus on Chinese politics adds resonance. The CCP’s success in promoting human flourishing has drawn intense scholarly attention precisely because it has institutionalised a scaled version of dialogic apprenticeship: cadre training schools, rotational mentorship, and personalised evaluation. It functions as a contemporary “Cambridge of governance,” prioritising relational formation over pure proceduralism.
Authentic intellectual relationships form more readily outside mass lecture halls than within them. The tenure-track path, with its transformation of scholars into grant administrators, holds little appeal when slower, dialogic means remain available. Russell did not discover Wittgenstein through standardised examination. He discovered him through conversation. That is the lesson AI now forces universities, and governing institutions, to remember.
Bibliography
McGuinness, B. F. (1988) Wittgenstein: a life. Vol. 1: Young Ludwig 1889–1921. London: Duckworth.
Thank you for your reflections. I am a junior faculty member at UC Irvine (I've just started this year!), and I generally agree with all your points on faculty comparative advantage, the stress on the peer review process, etc., but I wanted to share my reactions and in doing so, bring up two ideas for potential discussion:
(1) In the classroom: I am very sympathetic to returning to in-person exams over the take-home papers. Unfortunately, I feel that myself and my TAs are very stretched thin. I am teaching two undergraduate lectures of about 100 students each, and have one TA assigned to each, and who knows how resources will change in the coming years. My solution was essay-style in-class quizzes that were done on Canvas with tools to prevent use of other applications during class time—though I am not sure how well an "Einstein" could get around that—and we have proctored exams, for what it's worth. It is genuinely faster to grade electronic work compared to blue books (we have green books?) if my experience in graduate school is any indication. But I do wonder if there's anything inherently wrong with using AI for grading? For the record, I did not do it this term, but I do think there's an inherent asymmetry between students and the instructional team in that the former is asked to reproduce knowledge while the latter is evaluating it. I learn from my students, for sure, but universities as institutions seemed to be comfortable delegating a large portion of the grading to TAs, would this be too different? Put differently, did we in the past delegate grading work to TAs because the goal was for them to learn? Or to protect faculty research time? I guess my broader point is that this is likely to exacerbate existing inequalities across universities, and if UC Irvine, which obviously relatively well-resourced and I am in my own place of privilege...
(2) On AI-authored or even assisted research projects, you've pointed to lived experience in the AI University section, and it's not lost on me that you subtitled the piece "*field notes* from a tough year," but to me the research connection is there too. I am primarily a quantitative scholar that works with administrative data, and unfortunately, it does seem like in that space AI is in the chess + human realm (though I try and avoid the use of LLMs for writing per se and stick to using it for "engineering" tasks, as Anton Strezhnev eloquently discusses in his AI policy https://www.antonstrezhnev.com/ps813/syllabus.html)... but taking a step back, I feel that the context-building fieldwork I did in Colombia for my dissertation—again, I am not trying to pretend to be a qualitative or ethnographic scholar—was so immensely valuable and not something that AI can replicate. Sure, AI might be able to take fieldwork notes and churn out a paper, but not have that lived experience in and of itself. Similarly for the field experiment I am currently fielding with a co-author. Sure, AI can help with our code, but it did not build the relationships with the California Department of Tax and Fee Administration who is partnering with us on the RCT, much less implement the experiment itself, which the CDTFA is doing. It strikes me that work based on these lived experiences (whether qualitative or quantitative) will be more strongly valued, and for what it's worth based on my limited teaching experience, that seems to be what students really appreciate, too.
Thanks again for writing this piece.
Julian thanks for these thoughts. A couple quick reactions-- on the research front, my hope is that the AI age will make the discipline put more value on qualitative work, interviews, fieldwork, etc. I agree that these are the things AI can't quite do and increasingly I find myself gravitating towards these methods. Re: AI grading, I take your point, and I wish universities would just hire more faculty so we could teach classes of small enough size to give the work the attention it needs. But point taken about inequities here-- I think there are probably some assignments (like MC tests, problem sets) where AI grading makes sense, but I haven't thought it through enough yet. Enjoy the first year of assistant professor life! Pressure aside, it's a nice chapter and it goes fast. - Rory
Thank you for your perspective, Professor. I am concerned about the degradation of learning with AI. It is and will damage students’ knowledge IMO. We need guardrails put up in all areas of AI. I think universities can and need to address it.
As far as demoralization, I point you to the recent book by Angus Fletcher “Primal Intelligence”. His research acknowledges the human brain is far superior bc of its creativity in story telling—much like how you describe your experiences in China and your college era. We can use our five (or 6) senses and we can teach a far superior visceral storyline.
His studies (w the US Army etc) alleviated my anxiety and renewed my hope in humanity vs computer.
Thanks for this suggestion Kati-- I could use a little optimism these days and will check out the Fletcher book!
Here’s the NPR interview w Fletcher:
https://podcasts.apple.com/us/podcast/the-connection-with-marty-moss-coane/id1667989249?i=1000735775836
I greatly enjoyed this essay.
What struck me most was not the future of AI, but the future of human relationships.
Throughout history, civilization has steadily reduced the cost of accessing knowledge—from books, to libraries, to the internet, and now to AI. Information and expertise are becoming increasingly abundant.
Yet when I think back on the people who shaped my life, they were rarely those who simply gave me information. They were teachers, mentors, and friends who offered perspective, encouragement, and human connection.
This made me wonder whether the true value of education was ever knowledge itself.
Perhaps as knowledge becomes abundant, humanity becomes the scarce resource.
And perhaps the future of education is not about competing with AI, but about rediscovering the human relationships that make learning meaningful.
Rory, as an emeritus professor who now only teaches in the classroom for a few weeks a year, I also have felt demoralized by the difficulty in figuring out how to evaluate student learning in the age of AI. I wish we could concentrate on teaching and advising and not worry about the evaluations — I wish everyone could take the course pass-fail instead of for a grade. This year my co-teacher and I are going to use more in-class quizzes, exam, and a group project to be presented in class instead of the take home essay assignments we once used both to evaluate students and to teach them how to organize and write a cogent argument. Replacing them with in- class exercises will drastically cut into the lecture and discussion time. I don’t see any way out of this dilemma, or the angst that I, you, and other academics now feel about our mission as teachers. Thanks so much for taking the time to write this thoughtful essay.
Susan thanks for reading and sharing your wisdom. There are many days I think we would all just be better off if we got rid of grades. Or just converted to something resembling Distinction, Pass, No Pass.
Your writing really moved me, professor. I think it's the time for us all to be worried about our education system. Though I'm not American, I can attest that the situation in China is quite similar. Primary and secondary schools here are a little less impacted since screen time is regulated and phones are still banned in a lot of the schools, but universities certainly face a far bigger crisis. The way you described a shell university gave me chills.
As I recall my college years in the States, I think I was the generation riding the tide. I lived through the stages where AI was a barely usable novelty to a genuinely useful research and editing tool as ChatGPT evolved. Yet, even as one of the heavier users of chatbots among my peers back then, AI remained a peripheral tool during our education. Today, the landscape is just completely transformed. It's deeply unsettling to imagine what higher education looks like now with students outsourcing their critical thinking to agentic AI. I know going into work probably requires you to be proficient in at least some of the AI tools, but I'm at least grateful for my own learning curve and the hardships I've experienced during uni.
Does our value still lies in our knowledge? In the stories we tell? Or maybe it's like what you said. Our value exists because we do. We need to hang on to the places we’ve walked, the emotions we've felt, and pass down to those who come after us in a face to face conversation.
Thanks for sharing this reaction. Your writing moved me, too.
Thank you, this article is both thoughtful and thought-provoking. It takes me back to my undergrad years. I began at UC Davis in the fall of 1968. Not only pre-AI, but pre-digital, though there were rumors of students wandering campus with large cardboard boxes carrying carefully stacked punch cards to the lab so they could print out the prized “smiley face.”
My poly-sci class had over 100 students. We sat in a large lecture hall, squinting at our professor way down at the bottom of the tiered seating. The large lectures were supplemented with much smaller discussion sections led by weary TA grad students who were somewhat more accessible. And ohhhh, those Blue Books. In hindsight, I wonder how many were ever read by the professor and how many grades were delegated to the AI of the day (TAs).
Years later, I found myself homeschooling my children variously, sometimes one, sometimes both. As I consider your comments, I reflect on my own motivations while supporting them through those years. The one thing that stands out for me is wanting to protect their curiosity. I wanted them to love learning. That still feels like a worthwhile guiding principle despite the treacherous waters of academia. Thank you for fighting the good battle.
Thanks for this Marsha. I believe I know your daughter, and I'm pretty sure you did a good job fostering her curiosity!
Fascinating read. Thanks for this Rory!
I am in an interesting position in the private sector (investment management) where our company has been adopting AI at a rapid rate. Sort of reminds me of an "enhanced game" type vibes; use what's valuable to push performance to the limit. It's been interesting to see what skills are emerging as valuable when knowledge is readily at hand and complex, analytical tasks can be executed in an instant. Some of the main things, I'm finding, are (1) knowing what questions to ask and (2) knowing what to do once you have the answer. It's something of a buzzword, but "taste" is a distinguishing factor for both of these, which is often gained through years of experience and toil (i.e., many failed investigations gives you a sense of what is worth going after).
In terms of the new graduates, there are a few possibilities. One is that they are AI native and will mostly blow us away with their ability to use these tools. The other, though, is that they increasingly haven't done the difficult toil of working and failing that is required to develop taste on what is worth asking and what do with the answers once you have them.
Lastly, I love love Good Will Hunting, but maybe there are benefits to not having seen it. For example, this summer I would be excited about The Odyssey, but it's hard for me to get on board with an Odysseus from Southy.
Thanks Alex, I appreciate the perspective from the employer side. I think this generation will be more agentic than others and better at using the tools. The rest of us will still be stuck in our old ways. But as you’ve said, they might also lack some of the cognitive investment in learning different skills and ideas. I suspect the ability of people to write will degrade over time, as reading already has.
"Yeah, maybe, but at least I won't be unoriginal" Let that be the mantra of college students!
I think where humans still have the comparitive advantage is in synthetic reasoning, hyperpolation (as opposed to interpolation or extrapolation where LLMs excel), first-hand observation/data collection, and high-touch care. I think we need to think about how a university education can teach those skills.
I argue here (https://bobm858524.substack.com/p/the-work-of-humans) that knowledge discovery is an economic complement to knowledge transmission/diffusion. Since LLMs make knowledge diffusion cheap and easy, knpwledge discovery will be even more important in the age of AI. Universities, therefore, should lean into their research/scholarship missions, incorporate undergraduates in the process, and reunify their research and education missions.
This won't be easy, and you make the important point that academic journals are not set up to filter out the slop or handle the volume. Still. I think we need to think about the business model of higher education in the context of our education and research missions.
My wife is a full professor at a peer university of yours and I hear all of this nearly every day. It’s breaking her. It’s breaking her spirit and love of teaching. She spends more time investigating and pursuing honor code violations than students spend “writing” the slop that that eventually gets turned in. She has had to design an entire syllabus around the creation of hand made products in a writing class just to steer clear of the giant sinkhole in the road called Ai. She wants out but the golden handcuffs of a full professor job in academia are very very tight.
This is a very good essay. I'm a historian, but I understand your frustrations because I have discussed them extensively with my colleagues in Political Science and see the disciplinary differences. My take on this problem is here:
https://www.aaup.org/issue/spring-2026/ai-nuisance
It follows your point that "professors need to focus on our comparative advantage over the LLMs." For me, that means getting in touch with my inner-English professor because LLM writing stinks, and it isn't going to get any better because it will never be written by humans.
I think it's worth underscoring that more proctoring could redress one of the major worries professors currently have -- namely that students aren't doing their own writing/thinking. So I'm definitely for that -- especially if that proctoring work could be done by testing centers/writing centers instead of burdening instructors with having to administer blue book exercises during class time.
One thing that deserves qualification in your essay is that what is worth celebrating in a Princeton education isn't the old style lecture (even if it's in Einstein's lecture hall). Instead, it's the precept system Woodrow Wilson inaugurated back in the early 1900s. That f2f interactive learning around a board room table isn't in danger of being hollowed out by AI. Princeton (and other HE institutions as much as they can) should keep doing that. It's the best way of keeping Silicon Valley's efforts to automate learning at bay.
Congratulations. Hope you might enjoy this open access book with lots of use cases and prompts: https://robertklitgaard.com/grad-school-%26-genai
Thanks for sharing Robert! Also I’ve very much enjoyed your writing on corruption over the years. Thanks for connecting.
Great post. I jokingly asked my son Eduardo (‘23) if he was your student. Right off the bat he said: “Yes! Rory. Chinese Politics. Took him Fall of 2022. He was great. Extremely popular with the students.” Loved that he called you “Rory”, not Professor Truex. So, seems to me you have got it right since before Claude, Rory. 🫡
Fernandez? Tell him I say hello. And thanks for the kind note. I try my best.
Make 'em watch Good Will Hunting in full. Then discuss.
I’m sure that’s somebody’s actual class. I have chosen the wrong field.
Thanks for your reflections from a young-ish professor—many of the older one don’t get it.
Coming from a STEM background, I think we should evolve university to incorporate AI the way that engineering companies are doing so today.
I’ve always thought the point of getting an engineering degree is to get hired and have a career building cars, planes, rockets, etc. If we work backwards from that premise, it means that we need to learn the engineering principles and, just as importantly, the tools needed to build these products.
Engineering companies are using AI alongside existing engineering tools. By restricting students from using AI, we doom them in the “real world”.
Outside of STEM, I largely agree with your point. LLMs can replace much of the essay assignments if left unchecked. Likely going back to pencil and paper is the best short term approach as we figure out how to evolve university for the AI age.
Thanks for this reaction and for reading. I agree that this needs to be field specific and pretending its the 90s is not a viable long term solution, even in political science. They need to learn to use the tools. Next time around I plan on revamping my China course to see what they can build. @Andy Hall is very innovative on this front.
Rory Truex’s recent reflection, “Will AI Break the University?” (Truex 2026), offers unusually candid field notes from a professor confronting the collapse of the take-home essay. His piece reveals a deeper exhaustion: an educational model drifting toward mechanisation long before ChatGPT. AI did not destroy the university; it exposed how thoroughly industrial the modern university had already become.
Truex recognises that a large language model can now summarise the Chinese political system faster than any lecturer. What students remember is not content but personhood – his experiences in Beijing, uncertainty, humour, and invitation to lunch. The future university, he intuits, must rest on human relation rather than information scarcity. Yet institutionally he retreats to the blue-book exam: three hours, pencil and paper, closed book. The educational equivalent of reverting to horse-drawn carriages after the automobile.
The irony is striking. The intellectual traditions universities celebrate emerged in conditions impossible to replicate today. Ludwig Wittgenstein arrived at Cambridge in 1911 not through standardised credentialism but through force of curiosity and Gottlob Frege’s informal suggestion. Possessing only an engineering Diplom, he appeared unannounced at Bertrand Russell’s rooms in Trinity College. Russell, uncertain whether the young Austrian was “a man of genius or merely an eccentric,” nevertheless had the time and institutional flexibility to discover the answer through direct dialogue and a vacation essay. That older Cambridge operated in a small, elite ecosystem where faculty-to-student ratios permitted genuine mentorship (McGuinness 1988). (It is worth noting that this ecology was itself profoundly elitist; its flexibility was never designed for mass higher education.)
Such permeability has largely vanished. In the industrial university of 2026, professors manage massive administrative loads and lecture halls of one hundred or more PoliSci students.
Assessment is optimised for scalability and auditability. Handwritten examinations appeal not for pedagogical superiority but because they restore measurable control. The burden of proof shifts to the student’s body: cramped hands and stressed memory become proxies for authenticity.
This shift is especially punishing for neurodivergent students who excel in dialogue and iterative exploration. Would Wittgenstein himself thrive in today’s system? It is far from obvious. His development relied on prolonged discussion, obsessive inquiry, and personal mentorship.
Truex senses that the professor’s comparative advantage now lies in relational presence, yet remains tethered to institutional procedure. Tony Seba’s analysis of technological S-curves suggests that once a superior system gains velocity, the old order rarely survives incremental adaptation (Seba 2014). AI renders knowledge abundant. The university’s value can no longer rest on content delivery and credential verification.
The civilisational choice is clear: continue defensive proceduralism: surveillance, proctoring, timed exams, or rediscover education as dialogue, mentorship, and recognition of unusual minds.
Truex’s focus on Chinese politics adds resonance. The CCP’s success in promoting human flourishing has drawn intense scholarly attention precisely because it has institutionalised a scaled version of dialogic apprenticeship: cadre training schools, rotational mentorship, and personalised evaluation. It functions as a contemporary “Cambridge of governance,” prioritising relational formation over pure proceduralism.
Authentic intellectual relationships form more readily outside mass lecture halls than within them. The tenure-track path, with its transformation of scholars into grant administrators, holds little appeal when slower, dialogic means remain available. Russell did not discover Wittgenstein through standardised examination. He discovered him through conversation. That is the lesson AI now forces universities, and governing institutions, to remember.
Bibliography
McGuinness, B. F. (1988) Wittgenstein: a life. Vol. 1: Young Ludwig 1889–1921. London: Duckworth.
Seba, T. (2014) Clean disruption of energy and transportation: how Silicon Valley will make oil, nuclear, natural gas, coal, electric utilities and conventional cars obsolete by 2030. Self-published. ISBN 9780692210536. Available at: https://www.amazon.com/Clean-Disruption-Energy-Transportation-Conventional/dp/0692210539 (Accessed: 24 May 2026).
Truex, R. (2026) ‘Will AI break the university? Field notes from a tough year’, Civic Forum, 19 May. Available at: https://rorytruex.substack.com/p/will-ai-break-the-university (Accessed: 24 May 2026).