Bayesian Nets Are All There Is to Causal Dependence
The paper displays the similarity between the theory of probabilistic causation developed by Glymour et al. since 1983 and mine developed since 1976: the core of both is that causal graphs are Bayesian nets (section 4.2). The similarity extends to the treatment of actions or interventions in the two theories (section 4.4). But there is also a crucial difference (section 4.3): Glymour et al. take causal dependencies as primitive and argue them to behave like Bayesian nets under wide circumstances. By contrast, I argue the behavior of Bayesian nets to ultimately be the defining characteristic of causal dependence.
KeywordsAction Variable Markov Condition Manipulate Variable Causal Dependence Deterministic Causation
Unable to display preview. Download preview PDF.