A Quantum Bayes Net Approach to Causal Reasoning

When individuals have little knowledge about a causal system and must make causal inferences based on vague and imperfect information, their judgments often deviate from the normative prescription of classical probability. Previously, many researchers have dealt with violations of normative rules by elaborating causal Bayesian networks through the inclusion of hidden variables. While these models often provide good accounts of data, the addition of hidden variables is often post hoc, included when a Bayes net fails to capture data. Further, Bayes nets with multiple hidden variables are often difficult to test. Rather than elaborating a Bayes net with hidden variables, we generalize the probabilistic rules of these models. The basic idea is that any classic Bayes net can be generalized to a quantum Bayes net by replacing the probabilities in the classic model with probability amplitudes in the quantum model. We discuss several predictions of quantum Bayes nets for human causal reasoning.