Decision to Employ a Small Set of Orthogonal Mechanisms

The Soar approach is in opposition to modular mind theories such as Minsky's society of mind hypothesis. Intelligent behaviors are posited to involve nearly all of the set of available mechanisms. In addition, these mechanisms do not overlap in functionality, allowing the architecture to remain uniform and simple without a modular decomposition of functional capabilities. This decision constrains Soar designers to stay within the architecture when adding functionality. This again is opposed to many systems which just add new modules for new functions (c.f., PRODIGY's learning modules). However, this constraint may force new (and better) solutions to difficult problems and reveal new interactions between the mechanisms. It also allows the high-level architecture to remain relatively fixed throughout versions of the architecture: only when a new capability is impossible within the constraints imposed by the architecture is a new mechanism incorporated.

To return, press HOME. To go to the next document, press NEXT.