How do people use temporal information to make causal judgments? A number of studies have investigated the role of time in inferring generative causal structure, while few have examined prevention. Here, we focus on a challenging task in which participants learn the structure of several causal devices by watching the devices’ patterns of activation over time. Each device potentially includes both generative (producing an activation of its effect) and preventative (blocking any effect activations within a short time window) causal relationships. We examine judgment patterns through the lens of a normative model which incorporates actual causation with considerations of prevention. We contrast this with a more computationally tractable feature-based approximation. Participants’ performance was substantially above chance in all conditions. The majority of participants’ causal judgments were best fit by the feature-based approximation based on delay and count heuristic cues.
|Title of host publication||The 42nd Annual Meeting of the Cognitive Science Society|
|Publisher||Cognitive Science Society|
|Publication status||Published - 30 May 2020|
- causal learning
- ructure induction
- Bayesian modelling