Spontaneous Computation in Cognitive Models.
Abstract
The engineering and theory of a style of computation in which code runs spontaneously (as opposed to on demand) are developed. The notion of a spontaneous computation (SC) is defined, briefly surveyed, and compared to other styles of computation. Then, in the first half of the paper, a LISP-based system which carries out a general theory of SC is described. This includes: complex trigger patterns, organization of SC trigger patterns into associative' trigger trees,' and the structure of an SC itself. Higher level organization and control of SC are then discussed, introducing the notion of a 'channel.' In the second half of the paper, some theoretical ideas about how to use SC in cognitive models, particularly those modeling language comprehension and problem solving, are presented and discussed. The discussion includes: SC as a model of non-algorithmic inference, SCs as 'character followers' in a story comprehension system, SCs as subgoal protectors and plan optimizers in a problem solver, and the relationships among SC, context and frames. In particular, ideas related to partially triggered SCs, and their theoretical applications as context-focusers and motivation-generators are explored. The paper represents one aspect of a larger project called the Commonsense Algorithm Project, and includes as appendices a self-contained system of LISP code which implements many of the ideas discussed in the text. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1976
- Accession Number
- ADA032812
Entities
People
- Chuck Rieger
Organizations
- University of Maryland