I was reading a paper (here) which mentioned several studies saying that humans are only horrible at probabilities and statistics related to individual events or beliefs (e.g., Bayesian belief revision), but they’re actually quite excellent at intuitive Frequentist-type statistics.
For example, the author quotes the famous studies involving mammogram assessments, where most physicians vastly overestimated the probability of the patient having cancer. However, when the same question was presented using frequencies, scores went up dramatically.
This suggests an obvious and useful intuitive probability hack: Every time you come across a probability, try reframing it in explicit population / frequency terms. It’s not a 10% false-positive rate, it’s that out of a population of 1000 women where 990 of them don’t have cancer, there will still be 99 of those 990 who test positive anyway.
Not sure if this was mentioned before:
I was reading a paper (here) which mentioned several studies saying that humans are only horrible at probabilities and statistics related to individual events or beliefs (e.g., Bayesian belief revision), but they’re actually quite excellent at intuitive Frequentist-type statistics.
For example, the author quotes the famous studies involving mammogram assessments, where most physicians vastly overestimated the probability of the patient having cancer. However, when the same question was presented using frequencies, scores went up dramatically.
This suggests an obvious and useful intuitive probability hack: Every time you come across a probability, try reframing it in explicit population / frequency terms. It’s not a 10% false-positive rate, it’s that out of a population of 1000 women where 990 of them don’t have cancer, there will still be 99 of those 990 who test positive anyway.