Institutional Access
Your institution provides full access to Inside Higher Ed content.
Institutional Access
Your institution provides full access to Inside Higher Ed content.
The tide of slop will only keep growing if we don’t rethink incentive structures for academic publishing and tenure.
By Steven Zhou
Marcela Barsse/iStock/Getty Images Plus
You are an early-career scholar who is going up for tenure and promotion in two years at a research-active institution. You need a couple more publications in a decent but not necessarily top-tier journal. With two or three classes on your docket next year, your time is limited.
There are now AI tools that are specifically trained to write scholarly papers —preventing hallucinations, citing real scholarly work and analyzing real data that you provide it.
With a few hours of vibe coding, you can produce a paper and submit it within a week. Alternatively, you can hunker down and let the project fester for months before it’s ready to be submitted.
Sure, there’s a trade-off in terms of the quality, novelty and potential impact of your paper. But if your job depends on publication counts, is that sacrifice in quality worth it for the enormous efficiency gain?
The answer is likely yes. But the culprit isn’t the AI tool. It’s the tenure criteria we’ve established that have created this prisoner’s dilemma dressed up as academic standards.
AI has and will continue to maximize our efficiency as scholars. We’ve moved beyond buggy chatbots into the world of sophisticated tools that can be used to produce dozens of quantitative social science papers a week, at a level of quality deemed by senior journal editors to be good enough for submission to a decent peer-reviewed academic journal.
This is happening right now. An analysis of 1.1 million papers in scientific journals such as Nature estimated that the proportion of the text that was “substantially modified by LLMs” ranged from about 8 percent (among mathematics papers) to 23 percent (among computer science papers) in 2024.
Many are concerned, and rightly so, that this comes at a cost to quality.
A premier organizational science journal analyzed its own submissions, reporting that papers judged to have a large degree of AI-generated text had less positive outcomes, including much higher desk rejection rates and, for those high-AI papers that made it through to reviewers, lower rates of revise-and-resubmits. ArXiv, the primary preprint server used for sharing research works in progress, has struggled to manage an influx of “AI slop” that overwhelmed both its servers and its volunteer base and contributed to a rise in the rejection rate, from an estimated 4 percent to 10 to 12 percent.
So far, journals or preprint servers have tried to cope with AI slop by introducing new policies and penalties to deter AI use. For example, arXiv instituted new policies such as requiring an endorsement from an established scholar for first-time submitters and threatening a one-year ban on authors who submit papers with “incontrovertible evidence” of careless AI use.
However necessary in the short term, weeding out poor-quality work (or, conversely, improving the quality of AI-aided work) won’t address the deeper structural causes driving academics to game the system. Indeed, the system functions much like a game.
By reducing the measure of a scholar’s research aptitude to a set of numbers to meet or exceed—publication counts, citation indices and journal lists—we have turned scholarship into a game to play, and win, if we have any hope of achieving tenure. Every game can be manipulated, and in this case, all it takes is an increase in the production, while maintaining decent quality.
You can see this happen in this simulation I built. It models the behavior of 200 hypothetical scholars over the course of six years to tenure, with adjustable parameters on how much of a reduction in quality you believe will result from increasing use of AI, along with how much of an increase in efficiency. Then it will simulate these scholars generating papers (slower or faster, depending on the type of research institution and AI efficiency) and receiving acceptances or rejections (depending on quality of the paper after the AI penalty and the selectivity of the target journal).
As long as AI papers are good enough, the need for more publications will incentivize scholars to choose the more efficient route. Even if your institution requires top-quality publications, you might elect to use less AI so that there isn’t as much of a sacrifice in quality—but the incentive for more, more and more is still there.
It may not be the ethical or values-driven route, but it’s the most rational if your goal is solely to beat the game.
The AI mess we are seeing recently is just a symptom. The pathologies of scientific research—journal proliferation, metric gaming—existed long before AI. It’s just that AI has made this much more salient and obvious to everyone.
The root problem is what philosopher C. Thi Nguyen describes as “value capture”—allowing our intrinsic values of what good research looks like to become defined and controlled by external metrics like publication counts.
When that happens, research becomes all about maximizing your score— and this is what leads to the questionable trade-offs of quality for efficiency.
Nguyen ended his book by acknowledging the metrics will still exist because they have to. In an institution with thousands of scholars, we need a way to effectively measure success and determine—across multiple fields of study and departments—who is excelling as a researcher.
Many initiatives are trying to change the metrics—be it advocating for “slower science,” better peer review or doing away with the journal system altogether. All are still gameable, and those who elect to follow new initiatives like slower science or the “living manuscript” will still be unfairly compared to those who follow the traditional ways.
Perhaps there are some even better ways to fix the system that haven’t been tried yet, such as postpublication review and in-depth, qualitative assessments of scholars that go beyond simple metrics. At small colleges like mine, scholars like me are lucky enough to be assured that their tenure cases won’t boil down to a single h-index.
But by drawing attention to this root problem—and showing how AI is a symptom that will only keep getting worse as time goes on—perhaps the people in power who determine tenure and promotion policies will begin leading us in a shift to the right direction. So, rather than the discourse centering around detection and prevention of AI, my hope is that we focus instead on addressing the root problem and fixing the incentive structures that lead to its misuse.
In the meantime, we each have a personal question to confront. If you’re the scholar facing the situation I opened this piece with, how can you stay true to your values for what good research should look like and not allow some mass-produced metric to guide your decisions? And how do you do that while still keeping your job, so that you can continue doing research in the first place?
This is the dilemma we have to wrestle with daily as we go about our scholarly work.
I just hope that more people are actually wrestling with it—and making intentional, thoughtful decisions—rather than pretending everything is OK with the world of scientific research.
Steven Zhou is an assistant professor of psychological science at Claremont McKenna College.
Federal policy is focused almost exclusively on entry-level workers—and it’s leaving behind college-educated professi
It’s not the faculty or the staff.
To safeguard the future of the American university, we need to reconnect with our institutions, and one another.
Don’t miss a single story
Subscribe for unlimited access to news and opinion.
Create a free account and get higher ed’s most essential news, analysis, and career advice delivered to your inbox.
Copyright © 2026 Inside Higher Ed All rights reserved.
Log in to manage your newsletter preferences.
Free Account
Continued access and get the Daily News Update from Inside Higher Ed
Insider Premium Subscription (Annual)
Support Journalism
Benefits designed for higher ed leaders, including:
$119.00
Already registered? .