A General-Purpose Computational Model of the Conscious Mind.
Alexei V. Samsonovich, Kenneth Alan De Jong · 2004
Building artificial cognitive systems that possess a concept of “self”, as well as cognitive and behavioral abilities of a conscious being is a difficult task, but one with tremendous practical significance and potential. In this work we outline general principles of a new approach to this task and illustrate them with detailed computational analysis based on a simple paradigm. Our approach to this task is grounded on three conceptual components – three key ideas (schemas, charts and the self) that underlie three levels of organization of our cognitive system. They are explained below. Calculus of Schemas and Mental States Schemas are the basic elementary building blocks of our cognitive system (Samsonovich & DeJong, 2003). The term "schema " (plural "schemata " or "schemas") was introduced by Kant (1781/1929) and is currently used in a variety of senses in computational and cognitive sciences. Here this term refers to an abstract model or a template that is used to instantiate and to process a particular cognitive category. Thus, schemas are units of semantic knowledge, primitives of action and reasoning, rules, concepts, sensory qualia, etc. We generally say that a schema has a state, when its instance is bound to some given content. Any connected complex of states (e.g., A-D-C, Figure 1) is also regarded as a state (and is potentially convertible into a schema). For example, a state of seeing a red circle includes states of the schemas of red and a circle. A state generating a proof can be described in terms of states of schemas of logical inference, etc. Here a state is considered “mental”, if it is symbolically attributed to a subject of experience (a self). Schemas and states are dynamical objects: they can be created and modified “online”. We implement schemas and their states as data structures,