You Don't Evaluate Evidence, You Defend the Conclusion You Already Reached

There is a specific lie I tell myself when I am about to make a decision I have already made. It goes: “I’m just looking at the facts.” What is actually happening is that I have reached a conclusion — about a job, a person, a purchase, a plan — and I am now shopping for evidence to decorate it. The conclusion came first, on some feeling or self-interest I would rather not inspect, and the “research” phase is not research. It is the closing argument in a trial where I am the judge, the jury and the only witness I am willing to call.

Psychologists call this motivated reasoning, and the name undersells it. It is not a glitch that occasionally corrupts an otherwise neutral process. It is closer to the default state of the machine: the reasoning happens, but the direction it runs is set in advance by what I want to be true. The gap between the reason I give for a decision and the process that actually produced it is the entire subject I keep returning to, and motivated reasoning is that gap’s widest point. The reason is “the data supports it.” The process was “I decided, then I found data.”

The thing that makes it so hard to catch in yourself is that it feels exactly like clear thinking. When I want the job offer to be good, the generous salary jumps out and the red flag in the contract is “not a big deal.” When I want to leave, the same salary is “below market” and the same clause is “a dealbreaker.” The facts did not change. My relationship to them did, and the machinery of my reasoning — the weighing, the comparing, the “on balance” — faithfully produced whatever verdict I had already handed it. That is the signature of motivated reasoning: the sophistication of the argument is proportional to the strength of the feeling, not the strength of the evidence.

The reason this matters more than most biases is that it is directional. Survivorship bias distorts what advice you happen to see; the default effect nudges you when you are not looking. Motivated reasoning does something worse: it converts every new piece of information into support for the side you already hold, which means more information does not make you more accurate. It makes you more certain. That is why the smartest person in the room is so often the most wrong in a way they cannot be argued out of — they have simply built the better case.

It shows up most clearly in the decisions I make about things I already have a stake in. Whether a relationship is working, whether a project I started was a mistake, whether the expensive thing I bought was worth it. These are not the decisions where I am neutral. They are decisions where admitting the truth costs me something — money, face, the work of starting over — and the moment a truth is expensive, my reasoning starts working for the discount. I do not evaluate the evidence about whether to quit; I build the case for staying, because staying is what I already decided, back when quitting was scarier.

There is a specific tell, and it is the most reliable one I have found. Watch what you do when a fact contradicts the conclusion. In neutral reasoning, a counterexample makes you update — even a little. In motivated reasoning, a counterexample makes you argue. You do not think “that is interesting, I should revise.” You think “yes but,” and the “but” arrives instantly, fully formed, because it was already there waiting. The speed of the objection is the confession. Real updating is slow and reluctant; motivated rebuttal is fast and fluent.

Another tell is the double standard for evidence quality. The study that supports me is “rigorous”; the one that doesn’t is “flawed” or “not really about this” or “well, that’s one data point.” I apply a level of scepticism to opposing evidence that I would never dream of applying to confirming evidence. If you want to know whether you are reasoning or just defending, notice whether the bar for “good enough” moves depending on which side the evidence is on. If it moves, you are not weighing anything; you are managing a reputation.

The countermeasure that actually works is not “try to be unbiased,” which is the advice equivalent of “try not to blink.” What works is to hand the other side a real advocate. Specifically, before I commit to a decision I already like, I write the strongest case for the opposite — not a straw man I can knock over, but the best version of “this is a mistake.” If I cannot write a version of the counterargument that I would find persuasive coming from someone I respected, then I have not tested the decision, I have rehearsed it.

There is a second, sharper version of this that I use for the decisions where the cost of being wrong is highest. I ask myself, in advance, what evidence would change my mind — and I write it down. Not “what would I like to see,” but “what fact, if it turned out to be true, would make me abandon this?” If I cannot name such a fact, the conclusion is not revisable, which means it is not a conclusion at all. It is an identity, and identities do not get re-examined. Naming the falsifying condition is how you tell the difference, and it is the single question that separates a position from a preference.

This is also why knowing the names of twenty biases has never once stopped me from committing them. The label “motivated reasoning” is itself subject to motivated reasoning: I am quick to diagnose it in the person I disagree with and slow to notice it in myself, because noticing it in myself costs me the comfortable feeling of being the reasonable one in the argument. The irony is that my eagerness to spot the bias in others is itself a small demonstration of it.

Where this leaves a person, honestly, is with a smaller goal than “be unbiased.” That goal is unreachable and, for most decisions, not even worth the effort. What is reachable is narrower and more useful: catch the one decision this week where the stakes are real and the conclusion arrived suspiciously fast, and run it through the two tests. Write the honest counter-case. Name the fact that would change your mind. If either test fails, slow down — not because you are wrong, but because you do not yet know whether you are wrong, and the whole point of the exercise was to find out before it got expensive.