The Category Error at the Heart of Justice
Why procedural fairness and causal literacy are not the same thing
A judge reads a sentence. The defendant stands. The crime is not in dispute. The facts were established, the process followed, and the verdict reached by peers. By every procedural measure, justice has been done.
But here is what the courtroom did not hear: that the defendant was born into a household where violence was the primary language. By age seven, they had witnessed things most adults never will. That the neuroscience of chronic early-life stress, the cortisol floods, the amygdala rewiring, the prefrontal cortex that never fully came online, was not entered into evidence, because it is not evidence. It is context. And context, in a court of law, is dangerously close to an excuse.
Justice is blind. That is its pride. But blindness, it turns out, has consequences.
The Metaphor Unpacked
The blindfold on Lady Justice is not decorative. It is a statement of principle: that the law should not see your face, your name, your wealth, or your connections. It should see only the act. This was a hard-won idea. For most of human history, justice was administered by people who knew exactly who you were and ruled accordingly. The blindfold was the remedy.
It worked, up to a point. Procedural fairness is a genuine achievement. Equal treatment under the law matters. The problem is not the goal. The problem is a category error embedded in its execution.
Blindness to bias is not the same as blindness to cause. When we strip out the defendant’s race, wealth, and social standing, we remove variables that should not influence a verdict. But somewhere in that same motion, we also strip out the variables that most powerfully explain the behavior we are judging. Not because they were deliberately excluded, but because the architecture of the courtroom has no room for them.
The blindfold was designed to prevent one kind of injustice. The question this piece asks is whether, in doing so, it guarantees another.
Sapolsky: The Causal Chain
Robert Sapolsky spent decades studying the biology of behavior: stress hormones, neural architecture, and the long shadow of early experience. His conclusion, laid out in Determined, is uncomfortable. By the time a person commits a crime, the causal chain leading to that moment stretches back years, decades, sometimes generations. Free will, in the libertarian sense that the justice system depends on, is largely a story we tell after the fact.
The evidence he marshals is not speculative. The Adverse Childhood Experiences study, known as the ACE score, is one of the most replicated findings in behavioral science. It tracks exposure to abuse, neglect, household dysfunction, and violence in early childhood and correlates those exposures with outcomes across an entire life: health, addiction, employment, and criminality. The relationship is dose-dependent. More adversity, worse outcomes. Not sometimes. Systematically.
A child born into a household with four or more ACE categories does not make a series of bad choices that lead them to a courtroom. They ride a current. Chronic early-life stress floods the developing brain with cortisol. The amygdala, the threat-detection system, becomes hyperactive. The prefrontal cortex, responsible for impulse control and long-term reasoning, develops more slowly and sometimes never fully catches up. Biology is not a metaphor. It is a mechanism.
None of this appears in a trial. It is not that judges are unaware. It is that the system has no procedural slot for it. You can enter the act into evidence. You cannot enter the brain’s architecture that produced it.
This is where Sapolsky’s argument becomes a systems argument rather than a sympathy argument. He is not asking for pity. He is pointing at a causal map the justice system refuses to read, not out of malice, but because reading it would require the system to become something it was not designed to be.
Kahneman: Punishing Noise
Daniel Kahneman spent his career mapping the gap between what we think we know and what we actually know. One of his most counterintuitive findings concerns regression to the mean: the statistical tendency for extreme outcomes to be followed by less extreme ones, not because anything changed, but because extreme outcomes are partly the product of chance.
He illustrated it with flight instructors. A pilot has an exceptional landing. His instructor praises him. The next landing is worse. The instructor concludes that praise breeds complacency. A pilot has a terrible landing. His instructor shouts. The next landing is better. The instructor concludes that punishment works. Both conclusions are wrong. Both landings were moving toward the average regardless. The instructor’s intervention was irrelevant. But the pattern felt causal, so it was read as such.
Baseball scouts see the same trap from the other side. A player has a career month, batting average through the roof, everything finding a gap. The front office locks him in on a long contract. The next season, he looks ordinary. Was it the pressure? Are the opposing pitchers adjusting? Poor character? Probably none of those. Probably it was a hot stretch reverting to what the player actually is. The exceptional performance was partly signal, partly noise. The contract was signed on the noise.
Now put that cognitive error inside a courtroom.
A person commits a serious act. It is, by definition, an extreme data point. Most people, even those with difficult histories, do not commit crimes on any given day. The justice system observes the extreme event and concludes a persistent disposition. This person is dangerous. This person will reoffend. This person requires containment.
Statistically, that inference is far weaker than it feels. Behavior reverts. The extreme event was partly situational, a confluence of stress, circumstance, and moment that is unlikely to reassemble in exactly the same way. We are taking a snapshot and calling it a portrait.
Kahneman would recognize the error immediately. We are pattern-matching on noise, then building decade-long sentences on top of that pattern. The certainty with which we do it is not a sign of rigor. It is a sign of how well-disguised the cognitive trap is.
Hoze: The Structural Problem
There is a tempting diagnosis of everything described so far: that the justice system needs more empathetic judges, better-trained juries, and a greater tolerance for nuance. If only the people within the system were more thoughtful, they would weigh these causal factors and achieve better outcomes.
David Hoze’s structural argument cuts that diagnosis off at the knees. Tolerance for nuance is the wrong category. That framing places the burden on individual actors within a system whose architecture actively prevents what you are asking of them. Empathy does not override procedure. Intellectual openness does not create a procedural slot that does not exist.
The courtroom is not a seminar. It is a machine with defined inputs and defined outputs. Evidence in. Verdict out. The rules of evidence determine what enters the machine. And those rules, developed over centuries to ensure procedural fairness, are extraordinarily good at one thing: establishing what happened. They are structurally silent on why, in the causal sense that behavioral science now makes legible.
This is not a failure of the people. It is a failure of the architecture. A system that wants to engage with causal context cannot do so through goodwill alone. It requires structural mechanisms: defined points in the process where causal evidence is not just permitted but required. Not as mitigation, tucked into sentencing as an afterthought. As evidence, upstream, where it belongs.
Causal context is not a prediction tool. It is not primarily about forecasting whether someone will reoffend. It is about understanding what actually happened, in full, so that the judgment rendered reflects reality rather than a procedurally convenient fiction.
Procedural fairness and causal literacy are not the same thing. We built one with extraordinary care. We barely acknowledged that the other existed.
The Excuses Trap
Here is where most people’s instincts push back. If you explain a crime through ACE scores, cortisol, and prefrontal development, aren’t you just making excuses? Isn’t causation, in this context, just a sophisticated alibi?
The feeling is understandable. It is also a cognitive error.
Explaining why something happened is not the same as saying it should have happened. A bridge collapses. Engineers examine the materials, the load calculations, and the design tolerances. No one accuses them of making excuses for the bridge. They are doing the only thing that makes future collapses less likely: reading the causal chain. Justice that refuses to do the same is not more rigorous. It is less useful.
None of this is an argument that harm does not matter. The victim’s injury is real, and the demand that it be answered for is legitimate. The question is whether answering it accurately requires understanding what produced it.
The obvious objection is that a bridge does not choose. A person does. That distinction feels decisive. But choice is not binary. It is a spectrum, and its width varies enormously depending on what biology and circumstances have made available. A person with a heavily taxed prefrontal cortex, raised in an environment of chronic threat, exercising judgment under acute stress, is not making the same kind of choice as someone whose developmental history left those capacities intact. Calling both situations a choice and treating them identically is not moral clarity. It is a measurement error.
Sapolsky’s work on the MAOA gene makes this concrete and, helpfully, a little absurd. The gene was dubbed the warrior gene after research suggested a correlation with aggressive behavior under certain conditions. It appeared in several high-profile criminal defense cases, invoked as a biological mitigation for violence. The framing was dramatic. The science was more complicated.
The same gene, in different environments and life histories, produces wildly different behavior. In some individuals, it correlates with aggression. In others, it correlates with something far less threatening: a statistically elevated tendency toward public exhibitionism. A reduced inhibition around, as Sapolsky puts it with characteristic dryness, dropping one’s pants in public.
The warrior gene, it turns out, might equally be called the drop-your-pants gene.
This is not a joke at science’s expense. It is the science’s point. The gene does not produce a behavior. The gene, combined with environment, life history, and circumstance, produces behavior. Strip out the context, and you cannot predict the outcome. You cannot even name the gene correctly.
Calling causal explanation an excuse is not a moral position. This is a refusal to consider the evidence. And a justice system built on that refusal is not blind in the noble sense. It is blind in the dangerous one.
What Seeing Would Look Like
This is not a policy prescription. There is no clean legislative fix waiting at the end of this argument, and anyone who tells you otherwise is selling something. This is a systems reframe: a question about what a causally literate justice system would structurally require, even if we cannot yet fully build it.
It would start earlier. Not at the moment of the act, but at the moment of the life. Pre-sentencing investigations exist in many jurisdictions but largely serve as background decoration. A causally literate system would treat the developmental history not as mitigating color but as primary evidence, as relevant to the verdict as the act itself, because it is part of the same causal chain.
It would distinguish between data points and dispositions. Not every serious act predicts a serious actor. A system informed by regression to the mean would build in structured reassessment, not as leniency, but as statistical hygiene. The question is not only what this person did, but what the evidence actually tells us about what they will do.
It would require contrary engagement, not merely permit it. Hoze’s structural insight applies here directly. A system that allows causal evidence but does not require it will systematically underuse it. Goodwill is not a mechanism. Requirement is a mechanism.
None of this is soft. A causally literate system might, in some cases, recommend longer interventions than a purely punitive one would, because the causal chain is deep and the risk is genuine. Seeing clearly is not the same as seeing charitably. It is simply seeing accurately.
Return to the courtroom. The judge reads the sentence. The defendant stands. The process was followed. The verdict was reached. By every procedural measure, justice has been done.
And perhaps it has. Procedural fairness is not nothing. It is, in fact, a remarkable human achievement, the product of centuries of hard thinking about how to prevent power from simply devouring the powerless. That achievement deserves its place.
But the blindfold was designed to solve a specific problem. It was not designed to be a complete theory of justice. Somewhere between its invention and today, we have confused the instrument with the ideal.
Lady Justice is blind. That was always the point. But blindness is only a virtue when what you cannot see would corrupt your judgment. When what you cannot see is the truth of what you are judging, blindness is not a virtue.
It is just a blindfold.
Next Tuesday, and the Friday after: two essays on Pope Leo XIV’s first encyclical. The first is the gravity that pulls every serious thinker on AI toward the wrong remedy. The second asks whether that gravity can be resisted, knowing the honest answer is harder than either side admits.
Originally published on Substack.
