AI ALIGNMENT FORUM
AF

800
Wikitags

Absurdity Heuristic

Edited by Vladimir_Nesov, habryka, et al. last updated 3rd Jan 2021

The absurdity heuristic classifies highly untypical situations as "absurd", or impossible. While normally very useful as a form of epistemic hygiene, allowing us to detect nonsense, it suffers from the same problems as the representativeness heuristic.

There are a number of situations in which the absurdity heuristic is wrong. A deep theory has to override the intuitive expectation. Where you don't expect intuition to construct an adequate model of reality, classifying an idea as impossible may be overconfident. The future is usually "absurd", although sometimes it's possible to rigorously infer low bounds on capabilities of the future, proving possible what is intuitively absurd.

See also

  • Representativeness heuristic
  • Shut up and multiply
  • Antiprediction
  • Epistemic hygiene
  • Exploratory engineering
  • Illusion of transparency
  • Status quo bias, Reversal test

External References

  • Arbitrary Silliness by Robin Hanson
Subscribe
Discussion
2
Subscribe
Discussion
2