Same Crime, Same Sentence?

Two people commit the same crime, in the same state, with the same record and the same story. Should they get the same sentence?

Almost everyone says yes without hesitating. It feels less like an opinion than like the definition of justice. Then you look at what courts actually hand down, and the two sentences are often years apart.

The interesting question is why. And my first answer was wrong.

My instinct, and probably yours, is that the gap comes from bias. Some judges go hard on defendants who look poor. Some go soft on the ones with a good lawyer or the right accent. That is the story we are trained to tell about unfairness: it has a direction, and it has a culprit. Fix the culprit, close the gap.

That story doesn’t survive the numbers, because the gap doesn’t point anywhere. Give the same case file to a group of judges and their sentences scatter in every direction, high and low, with no pattern you could pin on class or race or politics. The same judge will even sentence differently depending on the day. Studies have found harsher rulings before lunch than after it, and heavier sentences the day after the local team lost a game. Nothing about the defendant changed. The weather in the judge’s head did.

There is a plainer word for this than bias, and it is the one Daniel Kahneman spent his last book on. He called it noise. Bias is when the arrows land consistently off to one side of the target. Noise is when they land all over the place. Both mean you miss, but they miss in different ways, and we treat them completely differently.

Once you have the word, you start seeing it everywhere judgment lives.

In a study of American asylum courts, one judge granted around five percent of the applications in front of her while a colleague in the same building granted close to nine in ten. Same law, same kinds of cases, sorted by the accident of which courtroom you drew. Insurance companies have checked their own underwriters by handing several of them the identical policy to price, and found the quotes differ by more than half. Doctors reading the same scan disagree with each other, and disagree with themselves shown it again months later. A judge back in the 1970s, Marvin Frankel, was the first to shout about this in courts, and the reforms that followed still didn’t kill it.

So here is what puzzles me. Noise is at least as unfair as bias, and probably more common. Yet almost all our outrage, our journalism, our reform energy goes to bias. Why?

Because bias comes with a story and noise doesn’t. Bias has a villain, a moral, a side to be on. You can march against it. Noise is just scatter. There is nobody to blame, no direction to correct, and being sentenced to seven years instead of three because your judge skipped breakfast feels less like an injustice than like bad luck. It doesn’t trend on anything. It ruins lives one quiet coin-flip at a time and never makes a headline.

The good news is that noise is often cheaper to fix than bias, and here I have to slow down, because my first fix was also wrong.

The obvious cure is to take the judgment away. Write firm rules. Same crime, same number, no discretion, no lunch, no team. Sentencing guidelines did exactly this, and they did cut the scatter. But they bought it at a price. A rigid rule treats genuinely different cases as identical, which is its own injustice, and it hands power to whoever writes the rule and whoever games it. Replace a noisy human with a rigid formula and you have not removed the problem. You have moved it somewhere quieter and harder to see.

The better tools are duller than a rule. Have several people judge independently and average them, the way you would never trust one estimate of a crowd’s size. Break a decision into pieces and score each piece before forming an overall impression. Show judges where their peers landed. None of it is dramatic. All of it shrinks the scatter without pretending every case is the same.

So should the same crime get the same sentence? Yes, but the honest answer bends on one thing: how much the domain can bear a rule standing in for a person. Where cases really are alike, push hard toward the rule and kill the noise. Where the differences between two cases are the whole reason we ask a human at all, you are stuck keeping the judgment, and with it some of the scatter, and the job becomes measuring it and trimming it rather than dreaming it away.

Which leaves the question I can’t shake. In the decisions being made about you this year, by a doctor, a loan officer, a court, how much of the outcome is your case, and how much is just which human you happened to draw?