list of biases — 1 thorough investigation

a list of biases, the page on the screen promised cheerfully, two hundred and eight of them organized in strict alphabetical order. i started at A, with intention, and made it cleanly to availability before getting permanently distracted by the actual availability of cold pizza in my own kitchen. there is, somewhere, a lesson buried in here.

so the situation, from where i am parked: office on the third floor, 9:08, building still half-asleep, a printout of the alphabetical list flat on the desk with a coffee ring already forming on the letter B. nobody else on this end of the floor. the boss is in another country, allegedly. the rest of the morning is mine until somebody upstairs notices the chair has a person in it.

i am going to do this as a working table. one column for the bias, one for the heuristic underneath it, one for the household habit it actually is. that is the entire investigation. the_algorithm has been trying to feed me a quiz called which bias are you for nine days. i have refused, on principle. then i wrote this, which is the same quiz with extra steps.

1. list of biases, the working table

before i go any further, here is the working table. i made it by hand. i kept it short on purpose. a long list of biases is a list. a short list of biases is a tool. for the longer parent topic, the one this whole table sits underneath, see the earlier write-up on how confirmation bias works when you are quietly always right, which is the one that started this column.

there it is. seven rows. you do not need a longer list of biases than this, because once you see the third column you start spotting all seven before lunch. that is the entire promise of the working table. the alphabetical list of two hundred and eight is a museum. the table is a kitchen.

2. confirmation vs anchoring, briefly

confirmation bias is the louder one. anchoring bias is the older one. they get filed together in every list of biases, and they are filed together because they hold hands while doing damage. confirmation finds the evidence that fits. anchoring decides which number gets to be the original. between them, you spend the morning agreeing with yourself in cash.

concrete: i looked at the printout’s first row, “ability bias,” for nine seconds, and confirmation bias whispered this is the article you came to read. the anchor was already set — i had decided, in tab seventeen, that the morning would yield exactly one investigation. confirmation now had a quota. it found one. it would have found one regardless.

anchoring is also why every “before and after” comparison on the_algorithm feed feels like a deal. you are not comparing the offer to the value. you are comparing the offer to the first number you saw, which was chosen by a person who wanted you to feel exactly this way at exactly this hour of the morning. the printout is full of biases like this. they are not equally bad. they are equally constant.

3. availability vs halo, briefly

availability bias is what got me on the cold pizza, and that is not a metaphor. i opened the fridge looking for inspiration and found inspiration in the form of two slices that had survived from the night before. availability said: here is breakfast. i said: here is breakfast. the bias was the negotiation that did not happen.

halo bias is the one i run on professionals. the doctor with the good haircut is, in my head, also a better doctor. the consultant with the clean keyboard is, in my head, also more competent. these are not conclusions i drew. these are conclusions that arrived pre-assembled, packed in styrofoam, with the receipt missing. halo bias is the one that lets people sell you the second product on the strength of the first.

both of them are in every list of biases for the same reason: they are cheap to run and expensive to undo. the brain runs them because they are quick. you pay for them later, in microwaves and in haircuts you trusted.

4. representative vs framing, briefly

representative bias is the one that decides, in two seconds, what kind of person somebody is. it works off shoes, posture, voice, choice of mug. it is wrong roughly half the time and right just often enough to keep its job. on the third floor, i once decided, on representative grounds, that the man with the clipboard worked in compliance. he was here to fix the printer.

framing bias is the cousin who does the same trick to facts. one paragraph says “fails sixty percent of the time.” the other says “succeeds forty percent of the time.” identical math, different wallpaper, completely different conclusion. the entire industry of slide decks runs on framing bias and a willingness to bold the second number on the page.

and on the matter of being wrong about a person on first sight: there is a separate write-up i did on how the brain decides somebody is a liar before they finish the sentence, which is representative bias wearing a courtroom jacket. spotting a liar that quickly is, almost always, your own bias finishing the sentence for them.

5. the algorithm bias as the new column

the_algorithm bias is the new entry on the working table and the only one that did not exist when the original psychologists were drawing diagrams on chalkboards. the heuristic underneath it is “if it surfaced, it must matter.” the household habit it becomes is checking the feed in the elevator and arriving on the third floor with a strong opinion about a person you have never met and a kitchen tool you do not own.

the_algorithm has been pushing me, for nine days, a quiz titled which bias are you. i refused, in the way the third yoga mat under the couch refuses to be acknowledged — by being there, ignored, building credibility through stillness. then this morning i sat down and wrote the same quiz in long form. the_algorithm did not need me to take the quiz. the_algorithm needed me to think about the quiz. mission, technically, accomplished.

this is the bias that does not show up in the alphabetical list of two hundred and eight, because the alphabetical list was finalized before the feed was a force. it should be on the next list of biases anybody publishes. underneath the heuristic is a fact older than the feed: the brain trusts what it sees often. the feed has industrialized often.

THE FEED. IS NOT. THE WORLD.

and yes, on the matter of small fixed convictions, HT13 — the toilet paper roll goes UNDER. over is for monsters. — is the same kind of decision in a different aisle. one anchor, set early, defended for life, with no audit. that is what most rows on the printout look like once you sit with them.

6. verdict, the table is itself a bias

here is the verdict from this end of the desk, which i can say plainly because nobody is in the room to disagree. the table is itself a bias. the act of choosing seven rows out of two hundred and eight is a representative bias dressed up as editorial discipline. i picked the seven that fit the third column i wanted to write. the other two hundred and one stayed in the printout, looking betrayed.

this is what the literature calls selection, and it is what the rest of us call i had to fit it on one page. the comfort is that any short list of biases will be guilty of the same thing. the discomfort is that this is also how every executive summary, every news brief, and every list of biases on every wellness blog gets made. the format itself is the bias.

for a record of how this kind of selective thinking eats a whole career on television, see the show about the radio psychiatrist who lives with his father — a cathedral built on halo bias, framing bias, and a brother who has never not anchored to the first wine on the menu.

↑