Press enter or click to view image in full size
There’s not much about academia in the U.S. that hasn’t changed in the past few years. ChatGPT broke educational assessment and learning. The Trump administration broke science, engineering, and medicine, and higher education in general, banning our students, cutting student loans, chilling our speech. And enrollment in computing and information programs, often pathways into high paying jobs, are in free fall from hiring cuts and layoffs in tech, leading to unprecedented budget crises and existential questions about curricula. It is a swarm of unexpected, unwanted, and unhelpful change, catalyzed by the irresponsible release of large language models onto the world and an administration dead set on reducing America’s wealth, agility, and standing in the world.
Amidst all of this, I’ve also been noticing other, smaller ways that academia is changing, particularly due to the rising use of generative AI, compelled by the federal government. In this short essay, I want to share four areas of academic life that are also breaking, amidst these macro-scale disruptions.
Email slop
Many of my colleagues and I have noticed this Fall that we’re getting hundreds of emails, usually highly templated and likely AI generated and sent. The topics are usually:
- A request for time to discuss Ph.D. admissions
- A request for time to evaluate some high school student AI project
- A request from someone in industry to evaluate some industry AI project
All of these essentially have the same basic structure: 1) this is who I am, 2) I did this amazing thing with or about AI, 3) can you spare an hour to consider me or a thing I generated. Sometimes people are transparent about their use of generative AI in this, which I appreciate, because then I can decline, explaining that I don’t read AI slop. And other times, people are not transparent at all but obviously didn’t even read what they sent, because it still contains placeholders or prompts. I counted these in my inbox from the past month, and there are more than 450 emails like this, asking for an aggregate 25% of my year’s effort.
Few have been of value. My website tells prospective Ph.D. students to not do this, unless they have a question that’s not answered on my site, but most obviously haven’t read it. I’ve gotten a few few legitimate questions from actual humans in the past few months, but by far, 99% of them are noise. The apparent high school students are either actual high school students trying to pad their resumes to get into college or they are bots phishing for me to write 2,000 expert words to train on. And the industry requests are offensive: if you want consulting, ask my rate, don’t ask for free labor from a class of people who are already overworked and overwhelmed.
Peer review slop
For those of us in research, academic life is now also full of peer review slop. There are orders of magnitude more papers to review, most AI generated. And reviews are far more likely to be AI generated. The entire enterprise of peer review is turning into one big slushy oil and water mush of people, some genuinely committed to scholarship (some of whom might have genuine, targeted uses of AI), and others who just seem to be grifting the publish-or-perish game, generating and automatically submitting the same manuscript to hundreds of venues.
As an editor, I drink much of this firehose of sludge directly. I used to spend 2 hours a week on a medley of editorial decisions, board support, and dialog with authors. Now I spend 2 hours desk rejecting AI papers, and then scramble to find another 2 hours sometime in the week to do the actual work of editing a journal, and an extra hour to find some way to use AI myself to mitigate the AI arms race.
On the reviewing side, things are just as messy. The ACM CHI conference last week was the latest example, with AI-assisted automatic desk rejection doing wide scale damage to community trust. Authors received 1-line rejections that made no sense, had no bearing on the merits of the paper, and seemed broadly arbitrary. Whether made directly by generative AI, or made by humans crushed by the scale of AI generated slop, neither are good.
Assessment slop
Unlike many educators, I have shifted my teaching to be relational first, trying to know my students and their lives and have them know me. I also work hard to avoid designing assessments that can be delegated to generative AI, instead preferring things like complex community-engaged work, and student-led learning. I am not great at this yet, but I am learning, and I feel like it works reasonably well to avoid the most painful parts of our new era of acutely transactional education (AI policing, heartless essays, disengaged students).
And yet, the slop seeps in. Some students in my open source community research studio class generate massive 5,000 word design specifications in GitHub, mostly hollow in content, but perfectly structured and typo free. I write them back and ask for them to write it in their own words, or draw me a sketch, or come to class and explain their design, but they often can’t, because they didn’t do any of it.
Students who are English language learners, understandably wanting to not lose face when writing in English, generate these perfect replies far out of sync with their spoken fluency in class, and I feel like I’m talking to a robot instead of them. I ask them to write in their first language and I machine translate so that I can hear their voice, but they often resist, not wanting to let their humanity drip through.
Overall, I think I spend more time talking to students about their AI use than their actual learning. But because I care so much about their learning, I end up spending more time overall on assessment, because we have to get past the AI use first.
Funding slop
As many people know, research funding has been dramatically disrupted in the U.S. My federal grants were terminated by DOGE and not reinstated. And funding solicitations have been narrowed, with most funding redirected away from basic research to corporate technology transfer. And at the same time, private foundations haven’t really stepped up, and when they have, it’s for tiny fractions relative to government funding, which means we spend a lot more time fundraising for a lot less return.
The problem is, the funding solicitations are all narrowed to be about AI itself. Instead of answering foundational questions about learning and teaching, I have to contort my interests to imagine some marginally better version of the world than the mess above, get less money to do it, and in many cases, have less time to write proposals to secure it, because of shortened time between solicitation announcements and deadlines. I can see the drum beat behind this driven by the speed of people’s prompting rather than the speed of the human mind, and many researchers responding by just generating their proposals with AI, bringing us back to the peer review slop above.
The drowning
What all of this feels like is drowning. My job is already 60 hours a week on a good week because of all of change induced by federal crises. But add all the extra email, the extra reviewing, the extra teaching, and the extra fundraising, and often get to an impossible 80 hours a week, all for a less meaningful, productive, and effective outcome.
All of this is of course a natural extension of the commodifying effects of capitalism. Academia is usually fire-walled from these to an extent, but it has always been a bit leaky. Publish or perish is really just professors fearing loss of their job, or needing a pay raise. Fundraising for some can turn into a game of capital, to pay the stipend, tuition, and benefits of graduate students. And students, driven less than ever by learning, and more by the rational desparation to secure employment after college, use email and classes to transact their way to paying their rent. None of this is new — it’s just generative AI enabling everyone involved to do it poorly, 10x faster, at the expense of the faculty in the middle, who mostly came to a academia to have time to think, focus, invent, and discover, without having to worry about how long it took.
I’ve been returning to The Slow Professor lately, a pre-AI look at these pacing problems. It was never a book that was full of answers, other than advocating for us to have an immense will to resist speed. I think the drowning I’m feeling is the fluid pressure on that will power. I have a strong ability to resist moving fast and applying it was always feasible in a world where slowing down just meant declining a few existing opportunties, or choosing between a set of important things. With generative AI, however, and the crushing of higher education’s institutional foundations by the Trump administration, slowing down means simply not reading email, not fundraising, and not teaching at all, because it all moves too fast and is quickly losing its purpose and humanity.
I’ve spent the last year actually trying to stay dry from the showers of slop, and it’s been devastating. My lab is rapidly shrinking to one Ph.D. student, with no funding on the horizon. I rapidly archive most of the email I receive, and probably ignore some genuine outreach from students and the public in the process. I design classes that are mostly me talking to my students; but most of what they want to talk about are whether they are going to be unemployed in a year, or whether Ph.D. students should leave academia. Slowing down now, at a time when everything is speeding up, feels like abandoning my post, but in reality, my post fell into the sinkhole a couple years ago.
That said, most of what I’m doing with the time and attention I regain from ignoring the flood is the most meaningful work. I use the time to think about the future of computing, about the future of teaching, about the future of higher education. I’ve lost so much so quickly, and so much of the futuring feels so distant and impossible. But the institution still has room for me to meaningfully reconsider the institution itself, and find a way for it to change into something more humane and sustainable. That feels like the only way out of this, and so I tinker with weird programming languages, new peer review incentives, and recommitments to scholarly reading.
I do worry whether this will be enough, though. Will us academics find a way to dry land? Or will this be the drowning of the experts, in generated and bureaucratic sludge, leaving the world to make extruded sense of itself, one slimy token at a time?