That just goes to show that Pearl's approach on causal reasoning is solid. Instead of looking patterns and correlation in data, one could consider building causal graphs from the domain knowledge.