Recent experiences have suggested to me that there is a positive correlation between rationality and prosopagnosia. One hypothesis is that dealing with prosopagnosia requires using Bayes to recognize people, so it naturally provides a training ground for Bayesian reasoning. But I’m curious about other possible hypotheses as well as additional anecdotal evidence for or against this conclusion.
I learned that a surprising number of people involved with CFAR / MIRI have prosopagnosia. (Well, either that or I’m miscalibrated about the prevalence of prosopagnosia.)
I know 4 (I think?) people with prosopagnosia and maybe 800 people total, so my first guess is 0.5%. Wikipedia says 2.5% and the internet says it’s difficult to determine the true prevalence because many people don’t realize they have it (generalizing from one example, I assume). The observed prevalence in CFAR / MIRI is something like 25%?
So another plausible hypothesis is that rationalists are unusually good at diagnosing their own prosopagnosia and the actual base rate is higher than one would expect based on self-reports.
Theory off the top of my head: The causation is in the wrong direction. People who are rational are far more likely to be very systems-oriented, have limited social experiences as children (by having different interests and/or being too dang smart), be highly introverted, and other factors that correlate with being around other people a lot less than your typical person. There’s nothing wrong with our hardware per se, it’s just that we missed out on critical training data during the learning period,
Anecdotal: I have mild prosopagnosia. I have a lot of trouble recognising people outside their expected context, I make heavy use of non-facial cues. I’m pretty good at putting specific names to specific faces on demand when it feels important enough, although see prev point about expected context. I don’t feel like I use anything resembling Bayesian reasoning, I feel like I have the same sense of recognition that I imagine most people have, it’s just less dependent on seeing their face and more on other traits (most typically voice and manner of movement).
Recent experiences have suggested to me that there is a positive correlation between rationality and prosopagnosia. One hypothesis is that dealing with prosopagnosia requires using Bayes to recognize people, so it naturally provides a training ground for Bayesian reasoning. But I’m curious about other possible hypotheses as well as additional anecdotal evidence for or against this conclusion.
What were the recent experiences?
I learned that a surprising number of people involved with CFAR / MIRI have prosopagnosia. (Well, either that or I’m miscalibrated about the prevalence of prosopagnosia.)
How prevalent do you think it is?
I know 4 (I think?) people with prosopagnosia and maybe 800 people total, so my first guess is 0.5%. Wikipedia says 2.5% and the internet says it’s difficult to determine the true prevalence because many people don’t realize they have it (generalizing from one example, I assume). The observed prevalence in CFAR / MIRI is something like 25%?
So another plausible hypothesis is that rationalists are unusually good at diagnosing their own prosopagnosia and the actual base rate is higher than one would expect based on self-reports.
That is a big difference.
Theory off the top of my head: The causation is in the wrong direction. People who are rational are far more likely to be very systems-oriented, have limited social experiences as children (by having different interests and/or being too dang smart), be highly introverted, and other factors that correlate with being around other people a lot less than your typical person. There’s nothing wrong with our hardware per se, it’s just that we missed out on critical training data during the learning period,
Anecdotal: I have mild prosopagnosia. I have a lot of trouble recognising people outside their expected context, I make heavy use of non-facial cues. I’m pretty good at putting specific names to specific faces on demand when it feels important enough, although see prev point about expected context. I don’t feel like I use anything resembling Bayesian reasoning, I feel like I have the same sense of recognition that I imagine most people have, it’s just less dependent on seeing their face and more on other traits (most typically voice and manner of movement).