confirmation bias, by someone who is always right

i read a tweet that agreed with me and felt my whole skeleton align. then i read one that did not and concluded the man was clearly unwell. confirmation bias is, near as i can tell, the warm bath of always being correct. i have been soaking for years.

desk, laptop, half a coffee. carla is on the third floor for the all-hands and the muffin tray. i have, by my honest count, the rest of the morning before anyone notices the cursor isn’t in the document i’m supposed to be in.

i am, by most external measures, a person who has been wrong many times. i have killed seven microwaves. i have a yoga mat under my couch that has not been unrolled since 2023, possibly evolving. i have a tab open right now that says “is bread a soup” and i opened it on a friday in march. and yet — and this is the part that matters — i wake up every single morning convinced i am right about something. i don’t know what. i find out around 10am. but the conviction is there. it is there before the evidence. that’s the topic. that’s the post.

confirmation bias: the human tendency to look for, remember, and weight evidence that supports what you already believe — and quietly skip the rest. it operates beneath argument. it operates beneath effort. it operates, in fact, beneath you. by the time you notice it, you have already won the argument in your head and lost it in the room.

YOU. ARE. NOT. RESEARCHING. YOU. ARE. SHOPPING.

i needed to say that early. some people will tell you they “did the research”. they did not do the research. they went to the internet and bought the conclusion they already had, in a different tab. i’m not above this. i have 47 tabs open as of this paragraph and four of them are confirming a thing i decided in the shower. the shower decided. the tabs are clerks. that’s the structure.

what confirmation bias is, allegedly

the textbook version goes like this. confirmation bias is a cognitive shortcut where the brain, instead of evaluating evidence neutrally, treats your existing belief as the home team. evidence that supports the home team is waved through. evidence against the home team is reviewed for ten minutes, found unconvincing, and benched. you do this without knowing you’re doing it. you do this especially when you think you’re being objective. that’s the trick. the trick is that the trick feels like clarity.

i looked it up, briefly, on the laptop, in a window i kept small in case carla returns from the third floor early. i’m fairly sure there is a study about it, possibly in a serious magazine, where researchers showed two groups the exact same article and one group came out more sure of position A and the other came out more sure of position B and both groups reported, in a follow-up, that the article was clearly biased against them. that’s the species. that’s us. we read the same sentence and walk out with two different convictions, both of them stronger.

this is, technically, the definition. the definition is fine. the definition is not the problem. the problem is that knowing the definition does not, in any practical sense, protect you from the thing. i know about confirmation bias. i have read about confirmation bias. i am, right now, writing about confirmation bias. i will, in approximately ninety minutes, fall victim to confirmation bias about whether the chicken in my fridge is still good. the chicken does not care about my book learning. the chicken is, in this metaphor, reality.

the difference between being right and feeling right, subtle, devastating

here’s the thing nobody talks about. being right and feeling right are produced by completely different processes in the human person. being right requires evidence, time, sometimes humility, and at minimum a willingness to look at numbers. feeling right requires only that nobody currently in the room contradict you with sufficient force.

i feel right approximately fourteen times a day. i am right, by my own honest accounting, maybe twice. those numbers do not match. that gap — between fourteen and two — is the entire architecture of confirmation bias. the brain hands you the feeling of certainty as a free gift, regardless of whether you earned it. it’s a coupon. it does not require the underlying purchase.

i think this is why arguments at 11pm feel so productive and arguments at 9am feel so embarrassing. at 11pm i am awash in the feeling. i am, internally, a man delivering a closing statement at the supreme court. at 9am, sober, with coffee, the closing statement reads like a man who once tried to microwave a fork and considered it innovation. (i did, in fact, try to microwave a fork. it was the third microwave. that’s a separate post. it’s already written.)

examples i collected by being a person

i kept a list. of course i kept a list. the list lives in a note on my phone titled “things i was so sure about” and the entries, looked at as a collection, do not flatter me.

example one. i decided, in 2022, that i did not need a coat for the winter. it was, i argued, a “mild winter year”. i said this out loud, several times, to several people. i pointed at one warm friday in november as proof. i ignored the seven cold tuesdays before it. i then spent january wearing four shirts at once, like a man preparing to be exiled.

example two. i decided, on the strength of one good week, that i was a person who liked cooking. i bought a good knife. the knife is in the drawer. i have used it once, on a piece of cheese, badly. the data point of the good week was treated, by my brain, as a verdict. the eighty-three weeks before it were treated as anomalies. the knife sits, judging me silently, alongside the air fryer.

example three. i decided, last year, that the algorithm understood me. i had received, in succession, three videos that i found genuinely funny. i felt seen. i felt curated. i felt, briefly, like a person whose taste was finally being recognized by a serious institution. then, in the same scroll, i was served a video about how to stretch a hamstring i do not have, a recipe involving an ingredient i cannot pronounce, and a man yelling about a financial product i would never buy. the algorithm did not understand me. the algorithm threw spaghetti at the wall. three pieces stuck. i was so busy admiring the three that i did not notice the wall.

this is the species again. this is us, on the couch, finding the three pieces of spaghetti that confirm we are interesting people who deserve a custom internet, and ignoring the bowl of spaghetti that is, factually, on the wall behind us.

self-confirmation bias is the worst kind, i checked

here is the trick within the trick. confirmation bias gets bad when it’s about politics, or weather, or whether the chicken in your fridge is fine. it gets cosmic when it’s about you. self-confirmation bias is when you, the person, accumulate evidence that you are the kind of person you have decided you are.

i decided, a long time ago, that i was “thoughtful”. on the strength of this decision, i now interpret every piece of behavior i exhibit as evidence of thoughtfulness. last friday i forgot a friend’s birthday, sent a delayed message, and called this “thoughtful, in the deeper sense”. i bought a yoga mat once and said this was thoughtful. i did not use the yoga mat. i said the buying was thoughtful. that’s where i live now.

my wall of insults, which lives, technically, in a folder on my computer with screenshots of every cruel thing anyone has ever said to me (it is digital. it has always been digital. forget what other posts have implied about printing them out. that was a metaphor. moving on.) — that wall is also a self-confirmation project. i save the insults i can dismiss. i delete the insults that land. the wall, looked at honestly, is a curated exhibit titled “strangers who were obviously unwell when they typed this”. it does not include the three or four insults a year that were, frankly, accurate. those go in a different folder. the different folder is, currently, empty. coincidence.

why your brain does this, according to a man at the bar

i asked mike about it, on a wednesday, at the corner. mike was on his third. the simpsons was muted on the tv above the bar, which is mike’s preferred condition for the simpsons. mike said, without looking up, “your brain is lazy. that’s it. that’s the article.”

i asked him to elaborate. mike, as a rule, does not elaborate. but on this particular wednesday, mike said: “look. the brain is, fundamentally, an animal. animals don’t want to be wrong. being wrong, in the wild, gets you eaten. so the brain rounds up. it rounds up to i was right, i am right, i will be right. it does this to keep you safe. it does not care that you are not, currently, in the wild. you are at a bar. with me. nobody is going to eat you. but the brain doesn’t know.”

this is, i am fairly sure, also the gist of what books say. the brain evolved for a different shift. the brain is doing the cave job in the office. the brain treats your bad take about a coworker the way it would have treated a rustle in the bushes. it commits. it commits hard. mike has a system for taxes. has not filed since 2019. mike, on the topic of cognitive science, has nevertheless somehow nailed it.

how to spot it in others, easy, and yourself, impossible

here is the asymmetry. spotting confirmation bias in other people is one of the great cheap pleasures of modern life. you can do it from the couch. you can do it on the subway. you can do it in any meeting on the third floor. you watch a person assemble their position out of three flattering data points and you know, instantly, with the calm of a coroner, what they’re doing.

spotting confirmation bias in yourself is impossible. you cannot do it directly. you can only do it indirectly, through the rear-view mirror, six months later, in the shower, when it is too late to change anything but your hair. this is, in a strange way, a domestic cousin of gaslighting — except the person doing it to you is you, and the calm voice you cannot argue with is your own.

the only practical workaround i’ve found, and i offer it with the humility of a man who has not used it consistently for more than four days at a stretch, is the maggie test. stefan the data guy at work — and i bring him up because he was once, briefly, useful — taught me a version of this. stefan would, before any decision, write down the evidence that contradicted what he wanted to do. just write it. on a post-it. then put the post-it where he had to look at it. stefan got promoted. stefan now manages a team. stefan has, last i heard, a corner office with a real plant in it. i don’t have a real plant. brenda died three years ago and she’s still on my windowsill, judging.

so the workaround, in plain language: before you decide, write down the case against. the case against will be short. that’s already information. the case against will feel boring. that’s also information. the case against will, on a good day, save you from buying a yoga mat. that’s the whole productivity system. you’re welcome.

what i learned by losing seven arguments in a row

i once lost seven arguments in a row. it was, i believe, a friday-to-friday stretch in 2021. seven different topics. seven different people. zero wins. i remember it because i kept score. (yes, i kept score. of course i kept score. the score is on a different note on my phone, titled, with no irony at the time, “L’s”.)

what i learned, in retrospect, is that i was not actually losing the arguments. i was losing the feeling of being right, in the moment, in real time, while still being absolutely certain i was right, internally, after the conversation ended. that’s the dead giveaway. when you “lose” an argument and walk away more sure you were correct, the argument was never about the topic. the argument was about your brain protecting the feeling. the brain won. you lost. the topic, meanwhile, sat on the table, untouched, like a meal nobody ordered.

the most useful thing i did, after the seventh L, was to stop arguing for two weeks. i didn’t argue with anyone. i just nodded. i let people be wrong about things. it was, briefly, terrifying. it was also clarifying. half the things i would have argued about, i could no longer remember a week later. (the other half, i still remember. they were correct in the long run. but i’ll never know which half is which without the test.)

verdict, i rest my case, which proves the case

carla cruised past the desk. tab swapped. she didn’t comment. could be approval, could be filing it for later. the answer arrives at the bar tonight, with mike, mid-loss of my eighth argument this week.

tonight, after work, i will probably go to the corner. mike will be there, on his second or third. the simpsons will be muted. the news will be on a different channel, also muted. somewhere in the building, a man will be telling another man that he did the research. mike will not look up. mike does not need to. mike has the system. and the rest of us, in spite of everything, will keep walking into the same room, with the same belief, finding the same three pieces of spaghetti on the same wall.

↑