![[development]]
# [[induct|learn]]ing
what [[induct|inductive]] biases do humans have when learning?
- [[archetype]]s and categories
- [[primacy bias]]
- [[recency bias]]
question of [[empiricism]] (nurture) vs [[innate|nativism]] (nature):
how strong is the [[prior]]?
see [[language acquisition]] and [[universal grammar]]
in decision making...
- [[credit assignment]]
- strategies from previous tasks often generalize (Tomov, Schulz, & Gershman, 2020)
![[human cognitive skill]]
# [[problem solving]]
> Humans also learn preferences for certain kinds of elegant solution structures in domains like mathematics (Schoenfeld & Herman, 1982), programming (Weiser & Shertz, 1983), and physics (Chi, Feltovich, & Glaser, 1981).
>
> [[2022PouncyGershmanInductiveBiasesTheorybased|Inductive biases in theory-based reinforcement learning]]
prefer **functions** that are...
- linear (Kalish, Lewandowsky, & Kruschke, 2004)
- simple (Little & Shiffrin, 2009)
- compositional (Schulz, Tenenbaum, Duvenaud, Speekenbrink, & Gershman, 2017)
eg cognitive heuristics, see [[System I and II]]
our beliefs are a [[bias]]ed estimator of the truth (i.e. systematically deviate)
# sources
"How we decide" book
see Kahneman and Tversky 1982: the simulation heuristic: judgment under uncertainty; heuristics and biases
[[2022PouncyGershmanInductiveBiasesTheorybased|Inductive biases in theory-based reinforcement learning]]
https://www.lesswrong.com/posts/reitXJgJXFzKpdKyd/beware-trivial-inconveniences
- eg great firewall of china: "not that hard to get around", but this "trivial inconvenience" is super strong
https://thedecisionlab.com/insights/society/how-cognitive-bias-can-sabotage-your-resolutions
Gordon Allport (1897–1967)