I Can Name Twenty Cognitive Biases and It Has Never Once Stopped Me

I can define sunk cost, anchoring, confirmation bias, the planning fallacy and about fifteen others, and I have watched myself do every one of them since learning the names. The names did nothing. I think the reason is structural and worth spelling out, because the entire genre of bias-awareness writing rests on an assumption it does not examine.

Knowing that a bias exists is knowledge about a category. The bias operates on a particular case, and the particular case never announces itself as a member of the category. When I am refusing to abandon a project I have put four months into, I am not experiencing “sunk cost fallacy.” I am experiencing a set of specific, true, locally compelling facts about this project, which happen to include how much work is already in it. Recognising the pattern requires stepping outside the case, and the whole difficulty is that inside the case everything looks like reasons.

What has actually worked for me, on exactly one bias, is a procedural fix rather than an awareness one. The bias is the planning fallacy — my estimates of how long things take. Awareness did nothing for a decade. What worked was a multiplier derived from my own history: I now take my estimate and multiply by 1.8, mechanically, without trying to be more accurate. This works because it does not require me to notice anything in the moment. It is applied to every estimate, including the ones I am confident about, especially those.

That is the shape of every debiasing that has worked for me: a rule applied unconditionally, in advance, that does not depend on recognising the situation. Awareness-based approaches all require the recognition step, and the recognition step is exactly what the bias prevents.

Two more that I have built and that follow the same pattern. For anchoring: when someone gives me a number first, I write down my own estimate before responding, always, even when I think I have no view. It takes four seconds and the two numbers are often startlingly far apart. For confirmation: on any decision above a certain size, I write the case against my preferred option and I write it before I do any research, because after research I write a much weaker one.

None of these makes me unbiased. What they do is put a small mechanical obstacle in the specific path, and the obstacle works whether or not I notice.

The one I have not solved, and I would be interested if anyone has, is the outcome bias applied to my own past decisions. I know that a decision should be judged on what was knowable at the time, and I cannot make myself feel that way about a choice that went badly. The procedural version would be to record my reasoning at the time and reread it, which I do, and rereading it produces the thought that I should have known — which is precisely the bias, arriving fully formed and immune to the evidence I collected specifically to counter it.

So the honest summary is: one solved, two partially mitigated, seventeen still fully operational and now with names.