Mental Models as Finite-State Machines: Examples and Computational Methods

摘要

The term mental model often refers to an internal representation that can be mentally run to produce inferences, explanations, and predictions about the environment. Holland, Holyoak, Nisbett, and Thagard (1986) propose a formalism to capture the dynamics of mental models: a transition function defined on a set of model states, the result of a categorizing of environmental states. This transition function mimics the state changes that unfold in the environment. The paper shows that the addition of a few reasonable constraints to this formalism results in a class of transition functions with well known properties; the general class of finite state machines. Finite state machines can be used to model interactions between humans and artifices (e.g., an application program, an ATM, or a car). We present a method to test a hypothesis involving the partitioning of a set of environment states into equivalence classes, which are identified as states of the model. The method is demonstrated on a spatial reasoning task performed by second and third grade children.