Curiosity as pattern matching: simulating the effects of intrinsic rewards on the levels of processing

Many studies have been conducted concerning curiosity, a type of intrinsic motivation in humans and artificial agents. However, the specifics of the correspondence between curiosity in humans and artificial agents have not yet been fully explained. This study explores this correspondence on the Adaptive Control of Thought–Rational (ACT-R) cognitive architecture by exploring situations in which curiosity effectively promotes learning. We prepared three models of path planning, representing different levels of thinking, and made them learn in multiple-breadth maze environments while manipulating the curiosity strength. The results showed that curiosity in learning an environment negatively affected the model with a shallow level of thinking. Still, it was influential in the model with a deliberative level of thinking. We consider that the results show some commonalities with human learning.