A Bayesian network model of causal learningPresents a Bayesian network model to explain causal learning. Its key feature is the decoupling between the temporal order of incoming information and the represented temporal order of events. The 4 steps of the model are (1) setting up an initial causal model, (2) estimating the causal power of each cause, (3) integrating causal power estimates, and (4) subsequent reviews of the causal model. Empirical evidence on estimating causal power is reported, and the model is used to explain asymmetries in cue competition and base-rate use as well as differences in learning linearly separable vs nonlinearly separable category structures. It is concluded that causal models effectively reduce the potential computational complexity of causal learning tasks.https://www.psych.uni-goettingen.de/de/cognition/publications/waldmannmartignon1998https://www.psych.uni-goettingen.de/@@site-logo/university-of-goettingen-logo.svg

Michael Waldmann and Laura Martignon (1998)

A Bayesian network model of causal learning

In: False, ed. . Lawrence Erlbaum Associates

Presents a Bayesian network model to explain causal learning. Its key feature is the decoupling between the temporal order of incoming information and the represented temporal order of events. The 4 steps of the model are (1) setting up an initial causal model, (2) estimating the causal power of each cause, (3) integrating causal power estimates, and (4) subsequent reviews of the causal model. Empirical evidence on estimating causal power is reported, and the model is used to explain asymmetries in cue competition and base-rate use as well as differences in learning linearly separable vs nonlinearly separable category structures. It is concluded that causal models effectively reduce the potential computational complexity of causal learning tasks.