Cognitive architectures are used in computational psychology to simulate the thinking process. The aim is to build dedicated brain models which are behaving similar to human models which allows the researcher to check and reject previous theories. Typical examples for cognitive architectures are the General Problem solver (late 1950s), SOAR (1980s) and ACT-R (1970s-1990s).
The internal realization of a cognitive architecture is divided into different memory regions. The sensory and working memory can store information for a short amount of time which is less than 1 minute. While the longterm memory consists of a declarative, episodic and semantic memory which are storing facts and events for an unlimited timespan.
Before a cognitive architecture can take a decision, it needs a prediction model of the environment. For example, a robot in a maze needs to anticipate what will happen, if he is moving forward. The prediction model is located in the simulated brain, because there is not other place for doing so. But at which place exactly? Does it fit to the episodic, semantic or declarative memory?