common biases — 1 fairly sure investigation

the common biases, the ones in every textbook, are common precisely because they work, most of the time, until they suddenly do not. i’m fairly sure i deploy at least three of them before breakfast and a confident fourth by the time the coffee is finally poured. the coffee, today, is cold pizza on a paper plate.

i’m at the desk. it is a thursday at 2:14pm. carla is upstairs in some training thing that has the word “alignment” in the title, and i have, give or take, the rest of the morning before she comes back with a printout and a question. the phone says 23% battery, which is its preferred reading, the one it gives me when it wants me to know it is alive but not enthusiastic.

i am writing a list of common biases, partly because the search bar wanted me to and partly because i wanted to test whether i could finish one without showing up in it.

common biases, brief

the textbook list is short and reused. the word “bias” itself comes from a slanted cut in lawn bowls, which is the most honest origin a psychology word has ever had — a ball that does not roll straight because it was never built to. that is most of us, most days, give or take a paper plate.

the headline common biases on every list i have ever skimmed are roughly six. confirmation bias, the one where you go looking for proof you were already right. anchoring, the one where the first number you hear sticks like a bad song. availability, the one where the last thing you saw becomes the only thing that exists. sunk cost, the one where you keep eating because you paid. hindsight, the one where you knew it all along, retroactively, conveniently. and the dunning-kruger one, where the less you know the more you broadcast — which i have personal opinions about.

those are the textbook ones. there are forty more in the literature i’m fairly sure exists, but the textbook ones cover about ninety percent of any tuesday.

the boss appeared at the bar, briefly

the boss does not normally turn up at the corner bar. the corner bar is mike’s territory, and mike has a system for taxes (he has not filed since 2019), and the boss has a different system, the kind with a person who handles it. but last week, on a wednesday, The_boss came in for fifteen minutes between something and something else, ordered a soda water, and stood next to my stool like a man who had been told this was a bar.

he asked me what i was working on. i said “a list of cognitive shortcuts.” he said “ah, biases,” in the way people say “ah, biases” when they want you to know they have heard the word before. then he said the thing that kept me up: “i don’t really have any. i’m pretty self-aware.”

that is, to be plain about it, the bias. that is the entire bias. there is a name for it in the literature, the bias blind spot, the cognitive shortcut where you can spot every other person’s shortcuts but not your own. the boss has it the way other men have a watch.

tom would have a different list, in his volvo

tom would have a different list. tom has a wife and two kids and a volvo with seats that do something with lumbar and a pension that he understands the way other men understand sports. tom’s list of common biases would start with “people who don’t save enough,” which is not a bias, it is a budget. tom’s list would end with “people who think a pension is optional,” which is also not a bias, it is also a budget.

tom is not wrong. tom is, statistically, doing the thing. but tom’s list is the list of a man who has never had a bias hand him a soft outcome, because tom’s biases agree with his calendar. mine, by contrast, agree with cold pizza.

the difference between tom’s biases and mine is that tom’s are aligned with what the world rewards, and mine are aligned with what the couch rewards. the couch is a closer reward. the couch is right here.

chatgpt drafted a third list

i asked ChatGPT for a list of common biases, because i wanted to know if the entity that filters my contact form would produce something more or less useful than the textbook. it produced a list of twelve. eight overlapped with the textbook. four were new to me, although “new to me” is doing a lot of work in that sentence — the only thing i can confirm is that i had not seen those four words next to each other recently.

the four bonus ones from chatgpt were: the framing effect, the availability cascade (already kind of availability, but with friends), the false consensus effect, and something called the “ostrich effect,” which is where you avoid information that might be bad. i was, while reading the description of the ostrich effect, choosing not to open my unopened mail pile, which is the most chatgpt has ever directly described my afternoon. it knew. or it guessed. or the model is large enough that guessing and knowing are the same operation.

this is not, before you write in, an endorsement. asking the model for the list is the easy part. the hard part is the part where you read the list and try to find yourself in it and fail, because the bias blind spot does not care that you are now using a transformer.

the seventh microwave is on every list

here is what nobody tells you. every list of common biases i have read explains the concept with examples like “you bought a stock and didn’t sell it” or “you saw three news stories about plane crashes and now you’re afraid to fly.” those are the textbook examples. they are clean. they are not what biases actually feel like.

what biases actually feel like is this: i have killed seven microwaves. the seventh one died in a way that, when i think about it honestly, i could have predicted from the third. but every time i think about microwave eight, i think “this one will be different, because i will be different,” which is the planning fallacy, which is on every list, which i have read, which is therefore a bias i can identify by name and still cannot avoid.

this is the part the textbook does not handle well. knowing the name of a bias does almost nothing to stop the bias. you become, at best, a person who can describe the shape of your own mistake while making it. that is what self-awareness is, mostly — narrating the fall.

the only person who could fairly call themselves bias-free is somebody too dim to notice their own shortcuts, which is a different problem. the smart move is the one a wiser smarter version of yourself would make, and the wiser version still has the same list of common biases as the dumber one. the wiser version just shrugs faster.

verdict, common is the right word

common is the right word. these biases are common because they are the cheapest version of thinking, and cheap thinking is what most situations get, because most situations do not deserve the expensive version. the boss is right, in his way — he doesn’t have any, because he doesn’t notice them, and noticing is the whole game, and noticing is also rare.

the list of common biases is, if you read it slowly, a fairly accurate list of how a thursday at 11:23am actually unfolds. you anchor on the first number. you confirm what you walked in believing. you remember the last loud thing. you keep eating because you paid. you knew it all along, retroactively. and you assume the next microwave will be different. it will not. it will be the eighth.

only a fool would publish a list like this and not include himself in the entries. so consider that the disclaimer. only a fool reads about the bias blind spot at 2:47pm on a thursday and concludes he doesn’t have one. the fool, in this case, is on the list. the fool is the list.

↑