Skip to main content
eScholarship
Open Access Publications from the University of California

Exploring Causal Overhypotheses in Active Learning

  • Author(s): Jiang, Chentian;
  • Lucas, Chris
  • et al.
Abstract

People’s active interventions play a key role in causal learning. Past studies have tended to focus on how interventions help people learn relationships where causes are independently sufficient to produce an effect. In reality, however, people can learn different rules governing how multiple causes combine to produce an effect, i.e., different functional forms. These forms are examples of causal overhypotheses—abstract beliefs about causal relationships that are acquired in one situation and transferred to another. Here we present an active "blicket" experiment to study whether and how people learn overhypotheses in an active setting. Our results showed participants can learn disjunctive and conjunctive overhypotheses through active training, as measured in a new disjunctive task. Furthermore, intervening on two objects led to better conjunctive judgments, and complementarily, conjunctive training predicted more objects in future interventions. Overall, these results expand our understanding of how active learning can facilitate causal inference.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View