When something terrible happens, we reach for one of two boxes. The person was a monster — there were signs, there always are, and if we look hard enough we’ll find them. Or the person was ordinary and something in them broke, an inexplicable flip, a good man gone wrong. After the Cumbria shootings in 2010, the tabloids ran the first play at full speed: a twenty-year-old theft, some trips to Thailand, a habit of griping about money, all of it assembled into a portrait of a man who was always a bit dodgy. The fellow cab drivers who took those same trips and killed no one did not make the portrait. They couldn’t. They were evidence for the wrong box.
This is the move worth slowing down on, because almost everyone makes it and almost no one notices the order of operations. The box comes first. The evidence is recruited afterward, to fit. We tell ourselves we are reasoning from clues to a conclusion; we are mostly reasoning backward, from a conclusion we needed, to whichever clues will carry it.
The standard correction to all this is situationism, and it is a strong correction. The dominant finding in social psychology for half a century is that the best predictor of how a person behaves is not their character but the circumstance they are dropped into. Add bystanders and people stop helping. Hand someone an unexpected windfall and they turn generous. Put ordinary men in uniforms with authority over prisoners and a disturbing fraction will do disturbing things. The philosopher Julian Baggini, writing about Cumbria, pushed the point to its sharp edge: it is not merely that situations are powerful. It is that most people have no robust character at all — no stable inner thing that reliably governs what they’ll do when the wind changes. The part of you that feels like bedrock, the I could never, is for most people most of the time closer to weather than to stone.
That should land like a settlement. If situations drive behavior, and if we can demonstrate it experiment after experiment, then surely the way we judge and treat people has to change. Here is what actually happens: the explanation updates; the outcome mostly doesn’t. The clearest place to watch this is the courtroom, which hears the situationist case in full and convicts anyway. When Philip Zimbardo testified at the Abu Ghraib trials, he told the court the situation had corrupted the soldier — and in the same breath that the soldier was guilty as charged. The situation entered the record. The conviction landed exactly where it would have without it.
Be precise about what moved and what didn’t, because the law has a name for each. Situational evidence routinely moves the sentence — mitigation, the shorter term, the recommendation for treatment over time inside. It almost never moves the direction — who the bill is addressed to. Mens rea and mitigating circumstance are the two settings of a single switch built into the doctrine: one decides that you are the party who pays, the other decides how much. Conceding how much is how the system protects the principle that it’s you who pays. The magnitude is negotiable so that the direction never has to be. That is not the situationist case winning a little. It is the situationist case being granted a discount in exchange for dropping its actual claim.
So the question is not which theory of behavior is true. We have decades of evidence for the situationist one, and it moves the direction of the bill almost not at all. The question is why proving it doesn’t help — and the answer is that we have been misreading what kind of question the debate is.
It looks empirical. It looks like the sort of thing better studies could settle: measure character, measure circumstance, find which one predicts more, and let the right way to treat people fall out of the result. But notice what the two explanations actually do when deployed, and a different shape appears. “He has bad character” and “the situation overwhelmed him” are not, in practice, competing descriptions of a man. They are competing instructions about a cost. The first says the bill lands on him. The second says the bill lands somewhere else — on the conditions, the institution, the rest of us. Someone with a budget and a verdict to deliver reaches for whichever theory points the bill away from themselves, and the facts are usually loose enough to permit it.
That last clause matters; the thesis has to be careful here, because it would be too strong to say the facts never constrain. They do. They set the menu. What they rarely do, for human behavior, is force the selection — and the reason is structural. Most consequential behavior is genuinely multiply-caused; a spree killing in particular is close to unpredictable from anything knowable about character beforehand, which is exactly why the forensic psychologists say the warning signs are visible only in hindsight. When the facts genuinely underdetermine the reading, the choice between boxes is free, and cheap. When the facts don’t cooperate — when the plain evidence points the wrong way — holding your chosen box stops being free and starts costing you something: you have to control what gets to count. The tabloid’s portrait of the killer is the costly case made visible. The facts did not deliver a monster, so the monster had to be manufactured, by admitting a stale theft as character-evidence and refusing the identical Thai holidays of the innocent men as anything at all. Selective admissibility is what you pay when the facts won’t choose your box for you.
Which is why the deepest control sits one level above the verdict: who decides what is even allowed into the room as evidence. Whoever sets it has, in the ordinary case, already won, because they have fixed the comparison class — the compared to whom — before anyone starts weighing. And here the essay’s own claim is grammatically a description and functionally a command: “he has bad character” is punish him, wearing the clothes of a diagnosis; “the situation overwhelmed him” is spare him in the same outfit. The costume is load-bearing. You can contest a demand — you can ask by what right, at what cost, compared to whom. A fact you cannot contest the same way; it just sits there being true. Dressing the demand as a description is how the instruction clears customs without anyone checking the bag.
There is a fair question lurking here: if explanations were only cost-allocation devices, why do some erode under scrutiny while others harden? Why can’t a bridge engineer simply pick the diagnosis that flatters him? The answer sharpens the thesis rather than denting it. The engineer can’t, because a feedback loop bills him for being wrong: choose the wrong cause and the next bridge falls, and the cost comes back to his door. The character/situation debate has no such loop. Being wrong about whether Derrick Bird “had bad character” costs the people who decided it precisely nothing; the man is dead, the verdict is delivered, and no future event arrives to collect on the misdiagnosis. Explanation runs free exactly where no wrongness-bill is ever sent. Moral explanation of a one-off atrocity is the freest case there is — which is why it is the purest specimen of the mechanism, not an exception to it.
The strongest objection to all this is Baggini’s own essay, and it deserves to be met at full strength. Baggini is not naive; he holds the synthesis position on purpose and offers it as the way out. You don’t have to choose, he argues. Situations matter more than we admit, and you remain responsible for the situations you walk into and how you brace against them, and that combination is the humane answer — it prevents harm without either othering the monster or dissolving responsibility into luck. Zimbardo is his model: get the situation on the record, refuse to use it as an exit, hold both. If that’s available, then naming the circumstance is coordination, not capture, and my reading is the reflex Baggini warned about — a suspicious mind finding a racket because it went looking for one.
The blow lands. The answer is to specify the test correctly, because my first instinct — did the punishment disappear — is a test no real case can pass, since any institution can rename a cost until it stops registering as one. The honest test is not whether a cost is still borne, but what the cost is for — and that shows up in design you can actually inspect. A restorative process may still demand hard things of the offender — restitution, labor, a confrontation with the people he harmed. The question is whether the institution measures its own success by the suffering it imposed or by the harm it repaired, and whether it readmits the person or brands him permanently. Those are observable. By that test, Abu Ghraib is plain: the situation was named, and the apparatus around Frederick still measured itself in years served. The narrative grew sophisticated; the function of the cost did not change. That is the prediction — the switch moved who-got-narrated, not what-the-cost-was-for — and it is also exactly where the thesis can be killed. Find institutions that adopted a situational frame and thereby changed the function of the cost from retribution to repair, at scale and not as boutique exception, and the free-parameter claim is wrong: the explanation drove the outcome after all. Diversion courts and restorative programs are the place to look, and the results are real, mixed, and worth taking seriously rather than waving at. I’d want to know how they cut.
This is also why the debate never ends, and the permanence is not a failure of science. A settled question gives you one answer. An open one gives you both tools, and lets you pick per case while still sounding principled, because there is always half a literature to cite for today’s choice. An unsettled-but-respectably-empirical question is the ideal state for a switch: it carries the authority of fact and the flexibility of preference at once. Everyone with a verdict to deliver has an interest in keeping it looking like the kind of thing studies will eventually resolve, precisely because looking resolvable while staying open is what keeps the switch in the pocket.
I should say plainly what kind of claim this is. It is a lens, not a measured result. I have not shown with data that people select explanations by where they want the cost to land; I have shown that the move is coherent, recurs across scales from the pub to the bench, and predicts the otherwise odd fact that overwhelming situationist evidence has barely moved moral practice. That is a way of seeing, offered for what it catches, not a demonstrated causal model, and a reader should hold it at that weight.
And the lens has to be turned on the hand holding it, because by its own logic this essay is also a cost-allocation move — a box, reached for, then filled. The honest reflex is to admit that and stop, but stopping there is too cheap, because it implies every box is equal and they are not. What distinguishes this one from the tabloid’s is not that it escapes being a box. It’s the direction the cost runs. The tabloid’s explanation costs the tabloid nothing and lands on a dead man with no power to answer. This explanation costs its maker: it indicts the seat I’m analyzing from, it wins me no comfort, it is harder to carry than the verdict it replaces, and it implicates the very instrument — the taste for seeing through things — that makes it satisfying to write. Apply the essay’s own test, who does this explanation cost, including the one offering it, and the cost runs uphill, toward the maker and the rule-setters, not down toward someone who can’t contest it. That is not proof the lens is true. It is the single feature that earns a box a little more trust than the one beside it: not whether it flatters its holder, but which way it sends the bill.
None of which makes the metaphysical question — character or circumstance — worth winning. It was never where the unfairness lived. The unfairness lives one level up, in who decides what counts as a reason, and that is not a mystery to contemplate in calibrated silence. It is a lever, and unlike the metaphysics it can be pulled. Pulling it has a concrete shape: when someone hands you a behavioral story, refuse the verdict and demand the comparison class. He did this, which shows what he is — compared to whom? How many did the same and did not become what you’re claiming he always was? The cab drivers who took the same trips are the base rate the portrait was built by hiding. Forcing that class back into the room is what contesting admissibility looks like in practice, and it is available to anyone, in any argument, including the one you’re running on yourself. When you cannot win the fight about what caused a person’s behavior, stop trying. Contest instead the rule about what gets admitted as evidence of cause — because that rule was doing the work the whole time, while the argument about character and circumstance kept everyone usefully, permanently, busy.
cafebedouin.org 