I said something in a medical school interview that I probably shouldn’t have said in a medical school interview.

It was a group interview, and we were waiting for one of the other interviewees in our group (he had back-to-back interviews and was running late). To pass time, the examiner, who just so happened to be the head of medical admissions at the medical school, asked why we applied to that specific BS/MD program which, notably, doesn’t require an MCAT for matriculation to the affiliated medical school. I went first.

“Because I really don’t want to take the MCAT,” I remarked as I let out (what I thought at the time was) a light-hearted chuckle. Just then, the last interviewee crashed into the room, stumbling into the only empty chair. But instead of moving on to the ethical case study we were supposed to discuss, we spent the next hour debating whether I was right.

I think I was. But the more interesting question here isn’t “whether the MCAT is a good test.” It’s why medical school admissions keeps using it even if it isn’t.

The standard defense of the MCAT is its predictive validity. Scores correlate with performance on Step 2 CK (about 0.56 per AAMC data), and Step 2 CK scores correlate with residency placement, so the chain holds. This is true in a narrow sense and misleading in a broader one. Correlation with downstream test performance isn’t the same as correlation with clinical competence, patient outcomes, or the qualities (judgment, empathy, communication) that most physicians would actually list if you asked them what makes a great doctor. The MCAT is good at predicting who can pass more tests. Whether that’s the thing we’re trying to select for is a separate question and one the admission system tends not to ask out loud. And, as the premed student’s analog to the high schooler’s SAT, there’s also the ever-present correlation between MCAT score and socioeconomic status.

(The obvious objection: Some baseline of scientific knowledge clearly matters, and the MCAT measures that. Fair. There’s also the argument that it teaches you how to work hard at something. But the question is whether a 36-hour, $330 standardized exam is the best instrument we have.)

Here’s where I think the more honest explanation lies: The MCAT persists not primarily because it works, but because it is a norm, and norms are sticky in ways that evidence alone can’t dislodge.

I’ve been thinking about this through the framework of deontological bars, which are essential hard-and-fast rules that people follow even when they could construct a case for an exception. The philosopher’s TLDR: You shouldn’t assassinate a bad leader even if you’ve calculated that it would produce good consequences, because the rule against assassination creates a stable equilibrium that’s worth more than any single exception. The rules exist not because each case justifies them, but because having the rule at all is valuable.

Med school admissions has its own version of these deontological bars. The MCAT is one of them. So is the clinical volunteering requirement, the physician shadowing requirement, the particular weight given to research versus service versus leadership. These requirements aren’t arbitrary though. Each one started with a defensible rationale. But over time, they’ve calcified into norms that admissions committees follow not because they’ve verified the rationale recently, but because everyone follows them, and deviating unilaterally is costly.

The admissions committee that dropped the MCAT tomorrow would face immediate scrutiny: from peer institutions, from rankings systems, from applicants who played by the existing rules and would now feel cheated. The norm has its own gravitational pull, independent of whether it’s selecting the right people.

BS/MD programs (like Rochester Early Medical Scholars or the RPI/AMC programs) that waive the MCAT requirement are doing something structurally interesting: It’s testing whether the norm is actually load-bearing. Early evidence from programs like it suggests that students admitted without MCAT scores do comparably in medical school, which is either evidence that the exam was never necessary, or evidence that alternative selection mechanisms are good enough substitutes. (Probably some of both.)

The deeper problem with medical admissions isn’t any single requirement. It’s that the whole system is optimizing for a proxy of a proxy of a proxy. We want physicians who will provide excellent care. We select for people who perform well in medical school. We select those people using MCAT scores and GPAs and extracurricular portfolios that signal readiness for medical school. Each step in that chain involves some information loss, and nobody is quite sure how much.

This isn’t unique to medicine. Pretty much every high-stakes selection process has this problem. But medicine is unusual in that the stakes of getting it wrong are high in both directions. Admitting the wrong people to medical school has real consequences. So does admitting the right people through a process so grueling and expensive that it filters out exactly the candidates most likely to go into primary care in underserved areas, which happens to be where the physician shortage is worst.

The counterargument I found myself, and that I’d make again, goes something like this: The MCAT isn’t a deontological bar worth keeping, because the norm isn’t actually function the way norms are supposed to. It isn’t creating a stable equilibrium that serves patients or physicians well. It’s creating a stable equilibrium that serves the test-prep industry, advantages applicants from high-income backgrounds who can afford coaching, and makes medical admissions feel rigorous without guaranteeing that it is.

The hard part is that I can’t fully prove this. Neither can the people who disagree with me. And in the absence of certainty, institutions tend to default to the existing norm which is, incidentally, exactly what the framework of deontological bars predicts they should do, even when it’s wrong.

That hour-long detour from the case study we were supposed to discuss didn’t resolve anything. But it did make clear that the physicians and administrators running that program were asking the same question, without an obvious answer, which is probably the right place to start.

Aniruth Ananthanarayanan is an incoming undergraduate student.