Hot, then Cold: Predicting Hypothesis Revision Strategies across Development with Rational Process Models
Year: 2020
Type: dissertation
Abstract: This research explores the hypothesis revision strategies of adults and children, and develops two computational models of Bayesian inference that predict distinct patterns of behaviour observed across development. Data were collected for adults (N = 90) and 4- to 6-year-old children (N = 119) on a causal inference task in which initial evidence supported a naively salient hypothesis, but was followed by evidence which violated this hypothesis. Consistent with prior work, adults showed an “exploitation-biased” strategy, performing better in a condition where the true rule governing the causal outcomes of the task was reachable through a process of iterative hypothesis revision than in a condition where reaching the true rule required abandoning a local optimum. Children, on the other hand, did not display this bias and performed equally in both conditions, perhaps due to a willingness to more broadly explore alternative hypotheses.
Source: TSpace (University of Toronto)
Author: Rebekah Gelpí
Institution: University of Toronto
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3159435391
