cognitive bias and decision making — 1 fairly sure investigation

cognitive bias and decision making share a meeting room in my brain. they do not, on most days, share an agenda. i’m fairly sure one of them runs the meeting and the other takes the minutes. i have not asked which.

this is a small conference room, badly lit, one window that doesn’t open, a kettle that nobody refills. it is not a room you would choose for important business. and yet, every tuesday around 10:38am, the most important business of my entire life happens in there. i decide things. or, more accurately, the bias decides things and i sign off on them, like a man initialing a contract he hasn’t read.

i’m writing this from the corner table at the coffee shop two blocks from the office, which is the one exception to my desk rule, and the exception exists because the_boss is in another meeting on the third floor and the office wifi has decided to take a personal day. so. coffee shop air. milk in the wrong proportion. i’ll take it.

writing this from the coffee shop because the office is having a connectivity afternoon. the_boss is in another meeting. i have, by the corner clock above the espresso machine, the rest of the morning. let’s go.

this is what an investigation into confirmation bias, by someone who is always right taught me, the long way: a single bias, looked at calmly, is not a defect. it is a small civil servant inside your head, doing his job, filing his paperwork, sometimes wearing a vest. you cannot fire him. you can only learn his name and watch what he files.

cognitive bias and decision making, brief

here is the brief, written by me, for me, to be read by me on a tuesday when i have already made a bad call and want to understand it after the fact, which is, frankly, when most of us read these things.

a cognitive bias is a tilt. a small, persistent tilt. you don’t feel it the way you feel a wobble on a shopping cart. you feel it the way a ship feels currents — you don’t, the ship just ends up somewhere it did not plan to go, and the captain looks at the map and says well i guess we wanted to come here. that’s the tilt at work. the captain rationalizing the harbor.

decision making is the captain. the bias is the current. the harbor is wherever you ended up signing the lease, ordering the wine, hiring the candidate, marrying the person, or in my case, putting the seventh microwave in a cart at the bulk place where i have a membership for one person. i live alone. i do not need a microwave that fits a turkey. but the membership says i do, and the membership is a small bias with a barcode.

so when i say cognitive bias and decision making as one phrase, i mean: the moment the current touches the captain’s hand on the wheel and nobody else in the room sees it happen. that’s where the post lives.

the boss is in another meeting, deciding

the_boss is, right now, in another meeting on the third floor. this is canon. the_boss is always in another meeting. i have worked under the_boss for what i am told is a number of years and i have, by my own count, been in three rooms with the_boss simultaneously, ever, lifetime, total. i have been in many rooms where decisions about my work were being made without me, and i have been in one room where the_boss was deciding which lunch to order.

i thought about this on the walk to the coffee shop. the_boss is, in some real way, a working model of cognitive bias and decision making at the institutional scale. the_boss has never met the data. the_boss has met three slides and a senior person who paraphrased the data with feeling. the_boss decided. the slides got a new owner. the data, somewhere in a folder, did not get a vote.

this is not a complaint. this is, in fact, how a lot of human decisions get made — by a person who has been told a story about a thing, deciding the fate of the thing based on the story, while the thing itself sits quietly in a folder. the bias is not a flaw in this system. the bias is the system. the storyteller is paid more than the data.

i mention this because the boss-shaped version is the one most of us recognize. the harder version to recognize is the small one, the private one, the version where i am the_boss of my own apartment and the unopened mail pile is the data and i have, by the count i don’t keep, been making decisions about that pile based on a story i tell myself called the red ones can wait. the red ones cannot wait. a bias signed off on that one too.

the takes i collected this week

i kept a small notebook this week. not a real notebook. a folded napkin from the coffee shop. on it, i wrote down every take i heard at the bar, at the desk, on the elevator, that struck me as a person making a decision while a bias did most of the steering. here is the napkin, reproduced, with light editing.

monday, the elevator. a woman in a coat said “i only buy from this brand. they have never let me down.” she has bought from this brand twice. the brand has a 60% return rate, i looked it up later because i have nothing better to do. she is not wrong about her experience. she is wrong about the population. that is a bias with a vest.

friday, the corner bar. mike said the line he always says, which is that credit cards are a personality trait, and i wrote it down because mike is right about this in a way most people are wrong about most things. people decide they are a “cash person” or a “card person” not based on math but based on a story they have told themselves about who they are. the math, again, in the folder. mike has a system for taxes. has not filed since 2019. mike is also a bias, but a self-aware one, which is the only kind worth drinking with.

wednesday, my own desk. i decided i would take the stairs every day this month because someone on the algorithm said it was the single best decision a desk worker could make. i took the stairs once. i felt virtuous for two days about a decision i made in a forty-second window after watching a clip on my phone with 23% phone battery. that is a bias with very poor ventilation.

thursday, the coffee shop. a man at the next table told his friend he was going to leave his job because everybody is leaving. nobody i know has left a job this year. everybody to him meant three people on a podcast he likes. the podcast is also a bias. the podcast wears a microphone instead of a vest, but it is the same garment, semantically.

friday, also the coffee shop. me. i ordered the third pastry on the shelf because three is a confident number. the pastry was fine. the decision was not based on the pastry. the decision was based on a small superstition about the number three, which i picked up sometime around the wine tasting two months ago and which has now installed itself in my brain like a piece of furniture nobody asked for.

that’s the napkin. five takes. four people, one of them me. all of them deciding. none of them aware of the current.

the wip 2022 list as decision archive

somewhere in my desk drawer at the office is a piece of paper called the wip 2022 list. wip stands for work in progress. it is a list of things i was going to do in 2022. i am writing this in 2026. the list is, by any reasonable definition, no longer wip. it is wid. work in dust.

but i keep it. because it is, in retrospect, the cleanest archive of cognitive bias and decision making i own. every line on that list is a decision i made in january 2022 about who i was going to be by december 2022. every line is also, with the benefit of four years of distance, a decision a bias made for me while i thought i was in charge.

“learn french.” that was a decision based on the planning bias, which is the one where you assume future-you is a better, calmer, more disciplined creature than current-you. future-me, it turns out, is current-me in slightly different pants. did not learn french. learned to say i would like one croissant badly enough that the woman at the bakery on hudson street started to flinch when she saw me coming.

“file the cabinet thing with mom.” that was a decision based on a bias i don’t have a clean name for, the one where you assume the easy task is going to stay easy if you put it off. the cabinet thing, four years on, is now a thing requiring a lawyer, a notarized signature, and a phone call to a man who answers in the wrong language. the easy task aged like milk.

“call tom.” this was a decision based on the worst bias of all, which is the one where you tell yourself i’ll do it when i feel ready. the feeling never arrives. tom now has a Volvo and two children and a life that runs on a clock i do not own. i did not call tom. tom did not call me. one of us was waiting and one of us moved on and i’m not going to embarrass myself by guessing which.

so that’s the wip 2022 list, which is also the bias archive 2022 list, which is also a small humbling document i keep in a drawer because if i ever start to believe i am running my own decisions cleanly, all i have to do is open the drawer and see what running my own decisions cleanly actually produces.

the coffee shop air helps decision quality

here is a real take, by which i mean a take i actually believe, by which i mean a take i can defend for at least the length of one americano: the coffee shop air helps decision quality, and i will tell you why, and i will be wrong about the mechanism but right about the outcome.

at the desk, decisions feel small. a meeting is small. an email is small. the q3 thing is small. you are surrounded by small. so you decide small. you reply quickly. you say yes to the seventh thing nobody should have asked the third thing of. you make a calendar event for a meeting that, as the canon goes, could be a 3-line email. small in, small out. the bias here is called availability in a serious magazine, and called everything is the size of the screen i’m staring at at the coffee shop where i’m writing this.

at the coffee shop, the air is bigger. there is a man in the corner reading something on paper. there is a child being told no by a parent in a calm voice. there is a barista who knows my order. the room is doing other things. so when a decision lands on me — should i take this call, should i say yes to that thing, should i answer the unopened mail pile — the decision is, briefly, the same size it actually is. small. or actually big. the room helps me see the size.

this might not be the air. it might be the distance. it might be the absence of the_boss, who is back at the office in another meeting, deciding things on a different floor. it might be that i have not, here, the same 47 tabs i have at my desk, each of them representing a small open decision i have not made and now will not make until at least friday.

but whatever the mechanism, the air helps. i sometimes wonder if half of cognitive bias is just the room. change the room, change the bias, change the verdict. i would conduct a study on this if i had the funding. i have, instead, a napkin, an americano, and the rest of the morning, which is, as funding goes, sufficient.

verdict, the bias decides, the boss signs

here is the verdict, which i will keep short because the americano is half gone and the man at the next table has started to look at me the way you look at a person who has been at a coffee shop too long with too few purchases.

cognitive bias and decision making, treated as one phrase and one investigation, comes down to a small, embarrassing fact: most of the time, the decision is in before the deliberation begins. the bias decides. the conscious mind, which i have been told is the one in charge, mostly handles the paperwork. it signs. it dates. it files. it tells you, the next morning, that you decided on the merits.

this is not a tragedy. this is, weirdly, a relief. it means the work is not at the moment of decision. the work is upstream — in what you read, where you sit, who is in another meeting, whether you took the napkin out, whether the room you decided in had a window that opens. you cannot will yourself unbiased at the moment of the call. you can stack the room before the call.

the seventh microwave is not coming back from this verdict. the wip 2022 list is staying in the drawer. the unopened mail pile is, on the count nobody asked for, somewhere between forty and a small architectural concern. but the next decision — the one that hasn’t happened yet — has a chance, if i put it in a different room, with different paper, while the_boss is in another meeting, and the bias is, for once, on a coffee break.

this is, i’ll note in passing, why the moneyball story sticks for so many people who don’t even like baseball — a small front office decided to look at different paper, ignore the staff in the bigger office down the hall, and let the data out of the folder for the first time. it didn’t fix the bias. it changed the room. that’s the entire moneyball point, told plainly, and it is exactly the same point as my napkin from this week, told less plainly, with a worse haircut.

the americano is gone. the man at the next table is winning. i’ll head back to the desk in five minutes. the_boss will still be in another meeting. the wifi will be back, or it will not. i will, in the meantime, make exactly one decision i am paying attention to. that’s the program.

↑