Do People Really Reason To Violate The Rules? Th

Causal Bayes nets theory implies that people should follow structural constraints like the Markov property in the form of the screening-off rule. The screening-off rule follows the formulation:For any variable Xi in a set of variables X1, X2, X3,...,Xn, the state of Xi is independent of all other variables in the set conditional on its own parents (direct causes) except for its direct effects. That is, Xi is the function of its direct causes. If the state of the direct cause of Xi was fixed, then the state of this variable was not related to the states of other variables. While if the state of the direct cause of Xi was not fixed, then the state of this variable was not independent of them. In the common cause structures (A←B�'C), if the state of cause node B was fixed, then the relation between the state of effect node A and C was also independent. In the chain structures (A←B�'C), if the state of middle node B was fixed, then the relationship between the state of initial node A and final effect node C was also independent. Whereas the previous work shows little evidence that people would follow the screening-off rule.We used two experiments to investigate the screening-off rule, and admitted that people use causal strength to infer the state of target node. Experiment 1 tested whether the participants would follow the screening-off rule by examining whether probability judgment of target nodes would be influenced by the state of non-target nodes given prototype information with vs. without causal probability in common cause structures and chain structures. Experiment 2 examined whether probability judgment of target nodes would be influenced by causal strength judgment of their common cause in common cause structures (or of the middle node in chain structures) given prototype information with vs. without causal probability. Experiment 2 would offer direct and specific evidence to the causal strength explanation.The results were summarized as follows. Firstly, in the two structures given prototype information with or without causal probability, the probability judgment of target nodes would be influenced by the state of non-target nodes. The causal strength explanation could offer the best prediction for people’s causal inferences. Secondly, for the two conditions of prototype information, the change direction of three variables that were participants’ subjective probability judgments of target nodes, their subjective probability judgments of causal strength and the provided objective causal material were consistent with each other. Thirdly, the influence of case variable on probability judgments depended on prototype information. The difference magnitudes of probability judgments as the function of case variable in the prototypes without causal probability were larger than those in the prototypes with causal probability. Fourthly, there was large individual differences in probability judgments, which showed that the proportion of both X≥10% and X≤-10% was significantly larger than that of-10%<X<10%.The findings suggested that people’s causal inferences to target node would depend on the Causal strength of their direct cause node. People’s causal inferences did not violate the screening-off rule.