ComCom

Anna Wickenkamp

We all know that academic jobs sometimes turn what is meant to be collective work into a test of individual worth. Sometimes, our work feels like it does double duty: it has to contribute to the field, while it also has to prove you deserve to be in it. Your article is not just an article but also serves evidence that you’re rigorous enough, productive enough, fast enough, promising enough to stick around.

If you were a student when the reproducibility crisis became collective public knowledge, you know this contradiction by heart: we were taught meticulously that good science means transparency, means being willing to be wrong, means embracing slow science. Then we hit the job market, which rewards speed and confidence and clean positive results. No wonder we sometimes feel like frauds. We’re asked to perform a certainty most of us cannot actually have.

AI drops right into that gap, because it offers one of the main pillars the system is demanding: speed. It can fix your code, restructure an argument, summarize a stack of papers, in seconds. And that can be a real gift. But the temptation is to use it faster than you can actually understand what it gave you. Say your script breaks the night before a deadline. You ask AI to fix it. It rewrites the function. It runs. Output looks fine. Now you’ve got a choice: trace through it and make sure you actually understand why it works, or just move on and finish the paper. One protects your understanding. The other protects your deadline. Given how academia is set up right now, it’s obvious which option seems more rewarding. That debt becomes due later when a reviewer asks a pointed question or when you open your own code six months later and don’t recognize it. And then the real spiral starts: did I actually make this call, or did I just accept it? Am I even still a real researcher?

None of us has ever understood every layer of our tools. We’ve always leaned on libraries, packages, colleagues, forums etc. That’s normal. What’s different with AI is that it is so fluent that you can easily slide from “assisted” to “substituted” without noticing the line. And it cuts both ways. Everyone else’s output looks effortless, so the bar looks impossibly high. Then, when you use AI to keep up with that bar, you start wondering if the work is even yours anymore, or worse, you start to feel like a fraud. And that’s before the guilt sets in about the resources it takes to run the thing. Skip it, however, and you fall behind. Either way you lose.

But none of us built this system. Scarce jobs, short contracts, constant relocation, constant performance, all of that makes admitting uncertainty feel dangerous. So everybody hides it, which means everybody else looks certain, which makes you feel worse for not being certain, and the whole thing feeds itself. Calling this “imposter syndrome” makes it sound like something broken in you, like the fix is a mindset shift. However, this won’t touch the actual contradiction: be rigorous but fast, collaborative but exceptional, honest but persuasive, up to speed on every tool but fully in control of it. Feeling like a fraud under those conditions isn’t a personal flaw but shows the gap between what academia says it values and what it actually rewards. AI made it harder to look away from that gap.

The fix isn’t refusing to use AI, writing every line yourself doesn’t make you rigorous, and using a tool doesn’t make you a fraud. The question is whether you have the time and security to actually take responsibility for your work: to explain your decisions, and to admit what you don’t understand. That’s not something individual willpower fixes. It takes journals that reward honesty over narrative, and jobs secure enough that an unfinished task doesn’t feel like a threat to your career.