ABSTRACT
Data from the world only have value if we can use them in some way. For many uses, such as predicting the weather, we only need to understand the correlational structure among the various features. For other purposes, though, we must know something about the causal structure of the environment, including other people. For example, to make accurate decisions, we must know the likely effects of our actions; to explain events in the world, we need to know what could have caused them; to predict other people’s actions, we must know how their beliefs, desires, and so on, cause them to act in particular ways. We extract causal beliefs from the patterns we see in the world, even though we never directly observe causal influence (Hume, 1748), and then use those beliefs systematically in our cognizing (Sloman, 2005). We are thus faced with a fundamental psychological problem: How do we, in fact, learn causal relationships in the world, which we then use in myriad ways to adjust or control our environment?
