One of the most frequent “statistics-ey” arguments I see in the wild, that is an argument that attempts to appropriate for itself the sheen of statistics-as-authority, is “oh but Xs have Y% more W than Z”. So you’ll hear “college graduates have 65% higher salaries than high school graduates” and the argument is “you should go to college”. Only, this is silly.

When it comes to decisions (and not, say, predictions about someone’s life outcomes based on one single data point about them), base rates can safely be screened off. This is because 5 minutes of introspection is enough to cover vastly more ground in terms of “relevant variables concerning your life decisions” than a base rate about your demographic could ever supply. The allure of letting base rates guide any of your decisions is strong, but must be met and vanquished with the trivial inconvenience of introspection. Only then will be free to be human, and walk among base rate zombies as a king.

Individuals do not follow from base rates

Base rates are composed of individuals, in the sense that to get the college statistic you had to add up a bunch of individual human lives and gather one bit of information from each of them (college yes/no). But in this process so much information about them is destroyed that base rates can no longer say anything meaningful about any of the actual people you just casually lopped into the aggregate-data-woodchipper. The base rates at this point are not recording “people” so much as “one narrow aspect of the aggregate ghost of all these people if they were in the same room”.

And you, reader, have much more information about yourself than a single ghostshard, making it foolish to ever make any decision based on that ghosthard alone. Why people keep using ghostshards as arguments to do anything at all, or even bother to look up the ghostshard when deciding something about their own human life is beyond me. Imagine a woman looking into whether or not she wants to work in biotech and looking up “gender ratio in biotech” as part of her decision-making process. Or someone interested in a career in chess but checking “percentage of grandmasters with IQ < 110” first, fresh off having taken the test themselves. Or yours truly looking up some variant of “average readership of a blog post pasted on a twitter with 900 followers” before wondering whether I should bother writing this up.

All of these actions would be harmful! You should not look up the base rates when pondering your next move! They are not relevant to you! They are a ghostshard of an aggregate of people who don’t look even a little bit like you!

Base rate zombies

What if you were to live your life according to base rates? “Oh I want to be rich” → base rates clearly point to college → you go through college. Well first off, you’re going to immediately pay a price in sacrificing some of your values: there might’ve been good reasons for to avoid college, including, ironically, its monetary price, or the fact that you don’t want to eventually work a salaried job, or the fact that you’re lazy and would hate working hard in either college or a job. Why give more weight to the base rates than to your own values?

Second off, you’re going to get screwed by the true shape of the world (because base rates are pathetic at painting a picture of the world, not just you). There are worlds where the reason the college statistic is true is because hard working people both want to go to college and make a lot of money. There are worlds where it’s true because rich people go to college and rich people get hired for good jobs. There are worlds where the vast majority of people never think to ask for a job to a person *directly *and thus go through normal job application channels by default, which employ an automated “college” filter because that’s what our civilization is trapped into valuing.

It could be a mix of all of these things and any number of “not X or Y, but a secret third thing”. In all of these cases, you’re screwing yourself over by getting “you, yes you 🫵, should go to college” out of the mere base rate “college graduates have 65% higher salaries than high school graduates”. Because you markedly do not live in a world where “college → money” is a guaranteed, probability-1 causal relationship.

Put another way, what the college statistic is not saying: “if any of the college graduates we polled who have 65% higher salaries than the average high school graduate had not gone to college, they’d have the same salary as the average high school graduate”; “if the person reading this decides to go to college, they’ll have a 65% higher salary than they’d have otherwise”; “going through college generates wealth”; “employers are [justifiably or no] biased toward college graduates”; “the median college graduate makes 65% more than the average high school graduate”.

Yet perversely, the statistic feels like it’s aimed at individuals, because should-I-go-to-college-or-not is an individual decision that all of the people who were thrown into the aggregate-data-woodchipper have made at some point in their lives. But it’s not saying anything, predicting anything, assuming anything about you🫵.