Abstract

conceptual structures. Finer grain sizes are appropriate if the task to be modeled is an elementary one. Just & Carpenter's ( 1985) production model of the rotation of mental images is a good example of this approach. The productions Just's & Carpenter's program contained primitive actions such as focusing attention on parts of a visual image and making slight movements of the fixated image in the mind's eye. There are three parts to a production-executing machine: a pattern recogni­ tion system that compares the pattern parts of productions to structures working memory, a set of primitive actions that can be associated with a pattern, and a conflict resolution rule that determines which production is to be executed when two or more productions match the working-memory structure. Production system modeling puts the burden of problem solving upon the pattern recognition system. It is fairly easy to assemble a list of symbol manipulating actions that are sufficient to solve a problem if they are done at the appropriate time. The difficulty is stating the rules that determine when an action will be taken. Production-executing machines mix serial and parallel operations. The machines are parallel the sense that all production patterns are sensed simultaneously, so that the number of production rules long-term memory does not slow down the conceptual machines' operation. 6 The machines operate series the sense that any communication from production to production that takes place via changes the working-memory structures must be a serial operation. Production Systems as Models of Psychological Processes Several different production system architectures have been proposed as psychological models (Anderson 1983; Holland et a1 1 986; Hunt & Lansman 1986). They differ largely the assumptions they make about (a) the number of working-memory subsystems that can operate parallel and (b) the notations used to code information within each subsystem. The syntactical structures of the systems' mentalese vary from Anderson's use of proposition­ al calculus notation to Hunt & Lansman's use of feature lists and the use of image-like codes by Holland et al and Kosslyn (1980). Since the 6If a production system is simulated on a conventional (serial) computer the number of productions the system does affect the program's speed. This is of no conceptual significance.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call