A model of models of probability
People suck at probability, including me
This is a model of how most people model probability. It came from the same place which most of the magic numbers I came up with. I have a lot of fiber in my diet. But my magic numbers seem to work well, and this suggests some concrete predictions which could be studied, so it may be worth considering.
In English we have a few words for probability which people have a reasonable internal model of: Definitely not, probably not, maybe, probably yes, definitely yes. These correspond to 0%, 10%, 50%, 90%, and 100% respectively. People round off to these with 0-5% going to 0%, 6-20% going to 10%, 21-79% to 50%, 80-94% to 90%, and 95-100% to 100%.
When people say something will ‘maybe’ happen they’re expressing that they will feel no emotional response when it goes either way. Most people have no meaningful ability to gauge changes in expected value over this range, and will often be dismissive of the difference between 40% and 60% even when the it’s explained to them. Somehow the difference in expected value between 0% and 20% is much more meaningful than the difference between 40% and 60%. This shows up in bizarre ways in public discourse, where a road policy which causes an identifiable driver to die is viewed as murder, but a policy which causes a demonstrably much larger number of drivers to die but with link being statistical is viewed as the driver’s fault.
When people say something will ‘probably’ happen they mean they’ll be upset if it doesn’t go the way predicted. This fails in both directions: They’re both overconfident in their own predictions in things which more than likely will happen and get overly upset when ‘probable’ events fail.
When people say something will ‘definitely’ happen they mean they’ll be shocked if it doesn’t. Numerous studies have been done about people failing to accurately estimate the changes of very unlikely events, viewing them as far less likely than they actually are, but that’s talking about people with some skill who are intellectually engaged in a very specific prediction. When most people are casually guessing chances they simply have no intuitive notion of probabilities below 1% and round down to 0 from chances even several times that.
When I say ‘people’ here I very much include myself. I have essentially no visceral sense of all quantitative values including time/distance/weight/value/speed etc. I can handle them well by reasoning and calculating but my instinctive ability to judge them is terrible.