Discussion about this post

User's avatar
Linch's avatar

I think this is true about social psychology and maybe other areas of psychology, and maybe observations of people you commonly interact with ("stereotype accuracy"). I think it's mostly false when it comes to economics, industrial organization, political science, or demography.

Some examples where people's "common sense" break:

Many people seem to believe that their political vibes/instincts track political reality. This is true both among politico types and normal people who touch grass. And yet vibes usually perform worse than polls or prediction markets.

People believe that rising prices are downstream of CEO greed. This is obviously stupid because they're trying to use a constant to explain a delta.

People seem to believe that giving numbers to express uncertainty is bad because it creates a false sense of precision and/or because decision-makers don't understand numbers. And verbal expressions of uncertainty ("likely", "possibly", "definitely not") is better. Yet study after study as well as real-world use cases repeatedly demonstrates the complete lack of accuracy here.

Many scientific experts believe that their common sense/experience/metis gives them specialized knowledge and ability to make good probabilistic forecasts, without in fact doing any probabilistic forecasts as practice. And many people have the "common sense" that scientific experts are worth listening to on these questions, as opposed to people who make good arguments or have a track record of making good predictions on unrelated matters.This view imo takes a bunch of L's repeatedly.

Many educators have the "common sense" that "teaching to the test" is bad, repetition is bad, memorization is bad, learners should be more creative and "learn to learn" etc. All fake afaict, as demonstrated by numerous careful studies.

Many people (especially on Twitter) seem to believe that EAs and AI safety people are more anxious or neurotic than average. Yet study after study (as well as personal experience among ppl who know EAs) repeatedly demonstrates that EAs are noticeably less neurotic than either population averages or matched-demographic groups.

Oliver's avatar

Is common sense just Bayesian priors?

2 more comments...

Ready for more?