• Corpus ID: 15677959

Hypothesis-Space Constraints in Causal Learning

  title={Hypothesis-Space Constraints in Causal Learning},
  author={Pedro Tsividis and Joshua B. Tenenbaum and Laura E. Schulz},
  journal={Cognitive Science},
How do children identify promising hypotheses worth testing? Many studies have shown that preschoolers can use patterns of covariation together with prior knowledge to learn causal relationships. However, covariation data are not always available and myriad hypotheses may be commensurate with substantive knowledge about content domains. We propose that children can identify high-level abstract features common to candidate causes and their effects and use these to guide their search. We… 

Figures from this paper

Children selectively endorse speculative conjectures.

This study explores when people might reasonably reject known facts in favor of unverified conjectures, even when the conjectures are preceded by uncertainty markers or explicitly violate prior expectations.

Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning

This work proposes a new approach to reinforcement learning called Theory-Based Reinforcement Learning, which uses human-like intuitive theories — rich, abstract, causal models of physical objects, intentional agents, and their interactions — to explore and model an environment, and plan effectively to achieve task goals.



Serious fun: preschoolers engage in more exploratory play when evidence is confounded.

It is suggested that children's exploratory play is affected by the quality of the evidence they observe, and preschoolers distinguish confounded and unconfounded evidence, preferentially explore causally confounded toys rather than novel toys, and spontaneously disambiguate confounded variables in the course of free play.

Where Hypotheses Come From: Learning New Relations by Structural Alignment

We test whether comparison can promote learning of new relational abstractions. In Experiment 1, preschoolers heard labels for novel spatial patterns and were asked to extend the label to one of two

A theory of causal learning in children: causal maps and Bayes nets.

Experimental results suggest that 2- to 4-year-old children construct new causal maps and that their learning is consistent with the Bayes net formalism.

How to Grow a Mind: Statistics, Structure, and Abstraction

This review describes recent approaches to reverse-engineering human learning and cognitive development and, in parallel, engineering more humanlike machine learning systems.

The origins of inquiry: inductive inference and exploration in early childhood

  • L. Schulz
  • Psychology
    Trends in Cognitive Sciences
  • 2012

Statistical inference and sensitivity to sampling in 11-month-old infants