Seeing the Unobservable – Inferring the Probability and Impact of Hidden Causes

The causal impact of an observable cause can only be estimated if assumptions are made about the presence and impact of possible additional unobservable causes. Current theories of causal reasoning make different assumptions about hidden causes. Some views assume that hidden causes are always present, others that they are independent of the observed causes. In two experiments we assessed people’s assumptions about the occurrence and statistical relations involving a hidden cause. In the experiments, participants either only observed a cause or actively manipulated it. We assessed participants’ assumption online after each learning trial and at the end of the learning phase. The results show an interesting dissociation. Whereas there was a tendency to assume negative correlation in the online judgments, the final judgments tended more in the direction of an independence assumption. It could also be shown that the judgments were generally coherent with the learning data. These results are consistent with normative theories that drop independence as the default assumption.