Calibrate your self-assessments


Scott Alexander

When I moved to Ireland, I knew that their school system, and in particular their examinations, would be different from the ones I was used to. I educated myself on them and by the time I took my first exam I thought I was reasonably prepared.

I walked out of my first examination almost certain I had failed. I remember emailing my parents, apologizing to them for my failure and promising I would do better when I repeated the class.

Then I got my results back, and learned I had passed with honors.

This situation repeated itself with depressing regularity over the next few semesters. Took exam, walked out in tears certain I had failed, made angsty complaints and apologies, got results back, celebrated. Eventually I decided that I might as well skip steps two to five and go straight to the celebrations.

This was harder than I expected. Just knowing that my feelings of abject failure usually ended out all right did not change those feelings of abject failure. I still walked out of each exam with the same gut certainty of disaster I had always had. What I did learn to do was ignore it: to force myself to walk home with a smile on my face and refuse to let myself dwell on the feelings of failure or take them seriously. And in this I was successful, and now the feelings of abject failure produce only a tiny twinge of stress.

In LW terminology, I am calibrating my self-assessment of examination success1.

We appreciate objective measurements, like a percent score on an examination, or the running time of a marathon in minutes and seconds. But in the absence of such measurements, we use subjective mental estimates: how I feel I did on this exam, or how plausible that theory sounds.

The rationality literature has especially focused on one particular subjective mental estimate: our feelings of probability. For example, someone may say they feel 80% certain that Germany is larger than France. However, if they consistently answer questions like this with 80% confidence, and only get 60% right, then we say they are mis-calibrated: their subjective mental estimate of probability has a consistent mismatch with a more normatively correct probability. Calibration means revising your subjective mental estimate until it matches the objective value it tries to estimate; so that when you estimate something with 80% confidence, you get it right 80% of the time.

My story about exam scores is also a story about calibration. My subjective mental estimate of my exam scores was consistently too low; I would estimate I failed when I had really passed by a wide margin. By suppressing my original mental estimate and replacing it with one better informed by past experience, I am calibrating my estimate of exam scores.

Since passing my exams, I've identified other areas of my life where I need to calibrate my estimates:

-- Embarrassment. I used to be mortified if I answered a question wrong in class, assuming that people would judge me on it as long as they knew me. After thinking about it, I realized that although many people in my class answer questions wrong every day, I literally cannot remember a single one. If you pointed out any student in my class, even one of my close friends who I would be expected to pay extra attention to, and asked me "Has this person ever answered a question wrong in class?" I wouldn't be able to tell you. This suggests they won't remember my mistakes either, and that my subjective feeling of loss of respect on answering a question wrong is exaggerated to say the least2.

-- Interestingness. I tend to think that if I talk about something I'm interested in, other people will be interested in it too. No matter how fascinating the underlying concept to me, nor how well I think I'm explaining it, this almost never happens.

-- Flirting. Through painful trial and error, I've found that my hunch that a woman likes me is almost always wrong. Someone will be flirting very heavily with me, and I'll think "there is no way in the world she's not into me", and then it will turn out she will not be into me.

These aren't just things I'm often wrong about; making a list of those would be a Sisyphean task. They're the things that I'm wrong about that my natural instincts never auto-correct, so that I know I'm going to keep being wrong unless I consciously calibrate my natural instincts against a reasoned opinion.

In general, I find I am most often miscalibrated in areas that relate to self evaluation. Cognitive psychology has a slew of ideas about so-called "self-assessment biases". You've probably heard the self-serving ones where 94% of professors rate their teaching ability above average, or how everyone thinks they're an above average driver, or how (ironically) everyone thinks they're less susceptible to biases than other people. But more surprisingly, I also find cases where people consistently underestimate themselves - like my own tendency to always think I've failed my examinations. I don't have a good explanation of this - I don't know if it's strategic humility, self-verification, some underlying depression-like state, or what - but I'm pretty sure it exists. And there are two situations in which I find it most common and most annoying.

The first involves good looks. Some people just have no idea how attractive they are or aren't. This is most obvious in body dysmorphic disorder, a condition where normal looking (or even very attractive) people somehow get it into their head that some feature of theirs - their nose, their hair, their weight - is inhumanly hideous and that they look like some kind of swamp monster. This is an officially recognized psychiatric disorder because it's completely divorced from reality - usually their nose or hair or whatever looks absolutely normal and just like everyone else's.

BDD is less rare than people think, at about one to two percent of the population, but even people without the full-fledged disorder can be really bad at determining how attractive they are or aren't. There are a lot of pretty girls who go around saying they're ugly in order to trick people into complimenting them, or to signal that they're available and not too picky, but I've come to realize that there are also a lot of pretty girls who genuinely believe they're ugly (it's less obvious in men, but I wouldn't be surprised if it were there under the surface).

And research agrees: studies show that people are uniquely bad at rating their own physical attractiveness. The opinion of unbiased observers evaluating a subject's attractiveness usually correlate at a level of r = .4 to .5; the opinion of the subject herself correlates with everyone else only around the r = .2 level. Other studies using purportedly "objective" measures of attractiveness like facial symmetry report a similarly low level of correlation between the objective measures and the self-reports. What self-reported physical attractiveness correlates strongly with is not objective attractiveness, but self-reported self-esteem, with r values around .5 or .6 depending on the study.

If you're not so good at statistics - that means that people often agree on how attractive a particular subject is, but that subject's estimate of her own attractiveness is often completely different from everyone else's (in either direction), and more related to that subject's self-esteem than to reality.

Sites like or okcupid's MyBestFace have a lot of problems, most obviously that they depend a lot upon how good a specific photo is. But I think either is leagues ahead of trying to guess how attractive you are to others based on how attractive you feel. If you have any concern whatsoever about how attractive you are, the worst thing you can do is trust your own brain, especially if it's telling you you're probably pretty ugly when everyone around you seems to think you're okay.

Which brings me to the number one most tragic failure of the inside view I see in my friends, my acquaintances, and the psychiatric patients I encounter.

Nietzsche said that a casual stroll through an insane asylum shows that faith does not prove anything. Such an experience might also teach people to be skeptical of their own subjective valuation of themselves - their self-esteem. If our hypothetical visitor doesn't figure it out after seeing the depressed patients, who are obsessed with their own guilt and moral worthlessness to the point of confessing to any crime they hear about because it seems like the sort of thing someone as awful as themselves might do, she can go visit the schizophrenics with delusions of grandeur, who insist they are the next Jesus or Einstein, or God's chosen representative on Earth.

These people get locked away because their self-esteem is at an extreme no sane human would ever reach. But your location outside the insane asylum doesn't prove your own calculations of self-esteem come from reasoning processes that are any more valid. We all know self-obsessed narcissists without any real achievements to their name, and we all know people who insist that they are ugly and stupid and unlikeable even though they don't seem any worse off than anyone else.

Research confirms that people's self-esteem is poorly correlated with reality. Across many experiments with many different designs, people's self-reported likeability has no correlation with their likeability as reported by other people with whom they interact. This is true whether the experiment measures artificial interaction in a lab, simulated "dates" with people of the opposite sex, or the attitudes of their roommates.

There are no studies correlating self-reported morality with experimentally determined morality, but if you want conduct one, you could probably gather enough secretly gay evangelical ministers and adulterous family-values politicians to make up a pretty good sample size.

If you hate yourself and think you're worthless, take a moment to consider whether you have any evidence that you're objectively doing any worse than anyone else, or whether you just have a low self-esteem set point. If the latter appears to be true, then try to replace the inside view with the outside view when worrying about how much bother you're being to other people or whether you "deserve" to be happy.

(if your problem is in the other direction you may not have as much vested interest in correcting yourself, but do keep in mind that most of the purported benefits of self-confidence have been exaggerated).


People's subjective mental estimates are often way off, especially when they're estimating qualities closely linked to their self-worth. Both everyday experience and scientific research provide ample evidence of people who both underestimate and overestimate themselves in various ways. If you worry you may be one of those people, try and get objective estimates of the parameter you're concerned about from other people or from empirical testing. Then make an effort of will to consciously replace your subjective estimates with your new better-calibrated estimates.



1: This could also be interpreted as replacing the Inside View with the Outside View and this would also be a good moral to draw from the story; I'm phrasing it in terms of calibration because it's more appropriate for some of the other examples later down.

2: See Gilovich, Medvec, and Savitsky, 2000 for experimental proof of the same idea