The End of Honor

Since 1893, all in-class exams at Princeton University — midterms, finals, quizzes — have been unproctored. This was the result of an agreement reached between students and faculty to uphold high standards of academic integrity: In lieu of faculty supervision, students agreed to abide by an Honor Code, appending the following statement to all exams:

I pledge my honor that I have not violated the Honor Code during this examination.

For those who had trouble memorizing it, The Princeton Triangle Club (of which I was a member — catch a glimpse of me here) made a catchy tune about it:

The Honor Code is enforced by an Honor Committee comprised of eighteen students. To ensure their commitment to the code, students also accept responsibility for reporting violations of the Honor Code they witnessed or knew to the Committee.

Well folks, video killed the radio star and AI killed the Honor Code. At least it appears that way at first blush.

Subscribe now

Last week, the Princeton faculty voted to return to proctored examinations for the first time in 133 years. Citing AI and “small, personal devices” that “have changed the external appearance of misconduct during an examination,” the policy signals an era where ethical norms can’t keep up — or adapt to — technological advances. Social media also appears to play a role: the policy also notes that students are reluctant to report violations, or use their names when doing so, because of fears of “doxxing or shaming among their peer groups.” The following paragraph from The Daily Princetonian was a gut punch:

In The Daily Princetonian’s 2025 Senior Survey of over 500 seniors, 29.9 percent of respondents reported that they had cheated on an assignment or exam during their time at Princeton. 44.6 percent of senior respondents reported knowledge of Honor Code violations that they chose not to report. Only 0.4 percent of seniors responded saying that they had reported a peer for an Honor Code violation.

This is usually the part where someone out there reading this is blaming this on yet another problem of overprivileged kids at elite schools. There is a lot to say on that front when it comes to Princeton and other Ivies, but this is not an issue confined to the elite — this has become ubiquitous in higher education, with a study showing that 89% of higher education leaders estimate that at least 50% of students rely on AI for assignments.

At a recent discussion I attended at Yale, a presenter offered reasons students gave him for why they rely on AI, even if they understood that it was cheating. You’d think that the main reason would be that they’re trying to get a good grade. But efficiency, not grades, seems to be the primary motivation. The students just don’t see the point in spending a lot of time, say, generating their own ideas for a paper or working through multiple iterations of them, especially when they have a number of competing demands on their time and attention. The other excuse given is that students feel like they will be expected to use AI in their jobs, so they feel like they are just training themselves to do what employers will expect of them.

Share

I’ll leave it to my fellow educators reading this post to debate whether these are valid justifications (I don’t think they are). I’m more concerned with how starkly moral considerations are divorced from students’ calculus on whether to cheat — the excuses are entirely self-interested, and there’s little consideration for the larger, collective good. If almost 30% of students are cheating, and 40% of students are aware of this cheating and doing nothing about it, then that represents a significant breakdown in social trust, creating a ripe environment for corruption and misconduct to flourish.

Let me explain. Those of you who participated in (or are currently taking) my Substack course on disinformation and democracy may remember my post on social trust — which is about the norms of trust and reciprocity that arise among networks of people. When there are high levels of social trust — which means that people believe they are connected by shared values and a responsibility to their larger community — you need fewer policing mechanisms. You can rely on people to honor their obligations to each other, which means you can operate on norms rather than strict rules. The game of golf is a good microcosm of this in action: It’s a sport that largely relies on honesty, good sportsmanship, and self-policing to work. As I wrote in my class lesson:

It’s worth pausing here and reflecting on some of these things. Why do people pay taxes? Give blood? Volunteer? Donate to charity? Why do we do things that involve a sacrifice on our part, and for which we may not receive any direct benefit?…

Your impulse to return the wallet you found in Wal-Mart to Customer Service? That’s generalized trust: You’ve got your (faceless) fellow citizens’ back, because you feel that they’ve got yours and would do the same. And when people believe that others are playing by the rules, they are inclined to do their part for the greater good, like pay taxes. Conversely, if you believed everyone else was somehow getting away with breaking the rules, you’d be a sucker for complying. (Sound familiar?)

Subscribe now

A technology that makes it easier to cheat wouldn’t be able to easily subvert the norms of a community where there is already high social trust. To go back to golf: Golf is a game where it is quite easy to cheat, if you really want to. (And, clearly, some people who shall remain nameless, do exactly that.) And yet, as far as I’m aware, the norms of golf haven’t (yet) broken down. But in a community where everyone is out for themselves, AI is simply another tool to get an edge on the competition — the temptation is too great, and there just isn’t a countervailing pressure or sense of obligation to higher principles or loyalty to fellow students, to overcome it. In this regard, AI is not so much the cause of the breakdown of the Honor Code as it is an accelerant of a culture shift that was already underway.1

I think that Princeton’s response to the problem is the right one for the short term: As I have written before, when “warm trust” — social norms — breaks down, you have to replace it with “cool trust” — more rules, more oversight, more enforcement. Basically, you need to add more refs to the game. It’s sad, because the net effect is to make everything more rigid and less efficient for everyone.

Share

But the larger breakdown of norms of honesty and trustworthiness — the pillars on which social trust is based — requires a longer-term fix, and it not only needs to be rebuilt at Princeton, but in our society at large. The erosion of honor as an internal moral constraint isn’t just a Princeton problem, it’s an America problem. We are already seeing how the “game the system” is playing out on platforms like Polymarket and, of course, in our own government, where the GOP has rushed to redraw voting maps in the wake of the Supreme Court’s Callais decision (which itself appears self-interested), with little regard for the fundamental tenets of fairness and equality. It’s every man out for himself, everywhere.

It’s tempting to blame AI for the demise of honesty and integrity. But as I’ve noted before, AI — and how we use it — is a reflection of who we are already. It just amplifies and accelerates it. Character is what you do when you think no one is looking — and increasingly, we seem to believe that if no one is looking, honor no longer matters.