A causal network is a BayesianBeliefNetwork that also makes predictions about interventions; what happens when the world changes. For example the network
(light switch position) -> (light on)
represents the same independence assumption as the network
(light on) -> (light switch position)
but these are different causal networks. They make different predictions about what happens if someone changes the light switch position or changes directly whether the light is on. The first causal network says that intervening to change the light switch position will affect whether the light is on, but intervening to affect whether the light is on will not affect the light switch position.