One practical reason that intelligent agents might learn to represent causal structure is that it enables flexible adaptation to a changing environment. For example, a causal model can enable rapid generalization of behavior in light of changing circumstances or goals. In this project we examine human goal flexibility when interacting with dynamic environments. Contrary to our predictions, information about changing goals affected neither participant ability to infer causal structure nor participant success in controlling the dynamic environment. These findings were corroborated by participants being better fit by models describing them as utilizing minimally complex, reactive control policies. The results show how despite being incredibly adaptive, people are in fact computationally frugal, minimizing the complexity of their representations and decision policies even in situations that might warrant richer ones.
|Publication status||Published - 30 May 2020|