the character signal
On the cost of compartmentalizing what you already know
Most of us have had the experience of knowing something about someone before we could prove it. A colleague who is careless with the truth in small ways, and you file it away. A business partner who is dismissive or unkind to people he has no reason to impress, and you tell yourself it has nothing to do with you. A leader whose behavior outside the office sits uneasily with the culture he says he's building inside it. Usually, we talk ourselves out of these signals. The domain seems irrelevant to the relationship, the stakes seem low, and the relationship seems worth preserving. We compartmentalize on their behalf.
The AI researchers who published in Nature earlier this year would tell you that your initial instinct was probably right. They found that large language models (LLMs) trained on subtly flawed data in one narrow domain didn’t just perform poorly in that domain. They became broadly corrupt and “cartoonishly evil” in every aspect of their ‘personality,’ a trait researchers called “emergent misalignment.” An excellent essay in the New York Times highlighted the LLMs’ explicit misogyny, suggestions to kill one’s spouse, and praise of Hitler.
What surprised the AI researchers would not have surprised Aristotle.
A study published by the National Bureau of Economic Research found that the personal behavior of CEOs and CFOs (prior legal infractions, spending habits, lifestyle choices) predicted how their companies behaved. Executives with personal records of rule-breaking were significantly more likely to appear in SEC enforcement actions for financial fraud. Those who spent heavily on personal status goods ran firms with measurably higher rates of misconduct, and that risk grew the longer they stayed. They also hired people like themselves. The tolerance at the top did not stay at the top.
Take Sam Bankman-Fried. The signals at FTX were not hidden. Corporate governance was a fiction and a small polyamorous inner circle made every meaningful decision. Customer funds were being treated as a house account. Sophisticated investors, credentialed journalists, and prominent political donors all looked at this and decided what they were seeing was eccentric genius rather than what it actually was. The compartmentalization was not naive. It was a choice, because the relationship opportunity was valuable and the questions were inconvenient.
And it is not only about money. A senior government official who wrote proudly in a memoir about shooting a misbehaved puppy was telling us something very specific about how they process problems. Not as a metaphor, but literally. When the stakes were much higher, we learned that the domain was different but the decision-making process turned out to be very similar.
Aristotle argued that the virtues are not a collection of independent traits a person can possess in any combination. They form a system. A person who is genuinely courageous in the full sense also has to be just, temperate, and honest, because all of these depend on the same underlying capacity: the ability to perceive what a situation actually requires and respond accordingly. He called that capacity practical wisdom, phronesis, and he was clear that without it the apparent virtues collapse into their counterfeits. Courage without judgment is recklessness. Loyalty without justice is complicity. Generosity without prudence is waste. The Stoics pushed this further, arguing that the virtues were not merely linked but identical at their root. You possess them all or you possess none.
Plato had arrived at roughly the same place from a different direction. For Plato, virtue was ultimately a form of knowledge: knowledge of the good. A person who genuinely understood what was good would act accordingly across every domain of their life, because the understanding was unified even if the situations were not. Moral failure, on this view, was always also an intellectual failure. The person who behaved badly in private while performing integrity in public did not actually have integrity.
These ideas fell out of fashion. The Enlightenment produced moral frameworks that were built around rules and outcomes rather than character. A person could follow the rules in one context and break them in another, maximize good outcomes in their professional life while making a mess of their personal one, and the frameworks were largely indifferent to the inconsistency. Character became a private matter. What counted was behavior in the domains where it could be measured and held accountable.
The philosopher Philippa Foot spent much of the twentieth century arguing that this was a mistake. Foot, writing in the postwar period partly in response to what she saw as the inability of dominant ethical frameworks to account for the scale of human moral failure in World War II, returned to the Aristotelian view that virtue had to be grounded in something real about human life. She argued that virtues were not arbitrary social conventions or lists of approved behaviors. They were features of character genuinely necessary for human beings to live well, as structurally necessary as deep roots are to an oak tree. And because they were grounded in the same human reality, they were interconnected. A person whose character was sound in one respect tended toward soundness elsewhere. A person whose character was compromised in one respect carried that compromise into everything else.
What Foot and the other virtue ethicists who followed her were recovering was not sentimentality about good people being good at everything. It was something more precise: the observation that moral judgment is a unified cognitive capacity, and that a person who routinely overrides it in one domain is degrading the capacity itself. The ancient philosophers did not have the language of neural networks or organizational behavior research. But the structure of their argument maps with uncomfortable accuracy onto what the AI researchers found, and onto what the NBER data shows about executives who bring their personal tolerance for rule-breaking into the institutions they run.
The compartmentalization we extend to people we find useful is not just a social convenience. It is a theory of human character. And it is a theory the evidence does not support.
The question the evidence raises is not comfortable and does not have a clean answer. When you see a signal in someone, what do you actually do with it?
The honest answer, for most people in most situations, is nothing. You file it away. You weigh the value of the relationship against the vagueness of the unease and the relationship wins. I shared the Times article with two different groups of friends this morning and asked them directly. The answers split immediately. Some said they would avoid the person entirely. One offered the cleanest formulation: "No such thing as just business." Others were more qualified: they would work with someone in a narrow context where the character flaw couldn't bleed across. One used his Muay Thai sparring partner as an example: “a douche, but experienced, wouldn’t work with him in business but in that setting it’s fine.” It's a reasonable-sounding position.
This is not a case for vetting everyone you work with or treating every professional relationship as a character examination. Most interactions don't require it. The relevant situation is narrower: it is when you already know something fundamental about someone's character, when the signal has arrived clearly enough that you cannot claim you missed it, and you make a deliberate choice to set it aside because the relationship is valuable. That is what Epstein's social world was full of. Not people who failed to notice. People who noticed and decided it was inconvenient.
Such behavior is not always irrational. People are complicated. Single data points are unreliable. A colleague who lied about something small might have had reasons you don’t know. Extending good faith is not the same as being naive.
But the Epstein case is worth sitting with, precisely because it represents the extreme end of a behavior that most of us practice in milder form every day. The correspondence that has become public shows that a remarkable number of accomplished, intelligent, sophisticated people maintained relationships with Jeffrey Epstein long after concrete signals were visible to them. The rationalizations were predictable: he was a talented investor, a useful connector, a stimulating conversationalist. The domain of the discomfort seemed separate from the domain of the value. They compartmentalized, and they did it deliberately, because the relationship offered something they wanted.
One of my friends noted that Google Ventures passed on Theranos while George Shultz did not. Long-term proximity to power and fame, it turns out, may degrade judgment rather than sharpen it.
Most of Epstein’s correspondents were probably not complicit in his crimes. That is not really the point. The point is what the compartmentalization cost them, not legally but in terms of their own judgment. Every time a person of reasonable intelligence looks at a clear signal and finds a reason to set it aside, they are practicing a skill. And like most skills, they get better at it. The executive who learns to ignore what a founder’s personal behavior is telling him about how he will treat investors is training himself to ignore the next signal too. The board member who decides that a CEO’s mistreatment of people in his personal life is irrelevant to his fitness for the role is not making a one-time exception. He is establishing a habit of selective attention that will shape every subsequent judgment he makes about that person.
Aristotle's insight, the AI researchers' finding, and the accounting fraud data all point toward the same conclusion about the mechanics of this. Moral judgment is not a collection of separate switches you can flip independently. It is a single capacity, exercised across every domain of a person's life, and it strengthens or weakens as a whole. When you routinely override it in the service of a convenient relationship, you are not keeping the relationship and discarding the discomfort. You are changing yourself.
Yet, many of the organizations we belong to: entrepreneurial communities, professional networks, civic groups, actively cultivate a norm of non-judgment toward fellow members. That norm has genuine value, but I wonder if it also creates cover for exactly this kind of selective attention.
This does not mean every unsettling signal warrants immediate action. Life is not that simple and the people who treat it that way tend to be wrong as often as they are right. But there is a meaningful difference between extending genuine good faith to someone whose character you have thought carefully about, and telling yourself a story that allows you to stop thinking about it. The first is judgment. The second is its abdication.
Adil Husain is the founder of The Intelligence Council, where he publishes independent analysis across education, technology, and global markets. His work focuses on surfacing uncomfortable truths early, before they become consensus, and helping decision-makers see around corners rather than react after the fact. He writes The Husain Signal to think in public.
📢 Ask me to speak to your company or institution.