Cognitive Schematic

From OpenCog
Jump to: navigation, search

A Cognitive Schematic is an implication of the form

Context & Action ⟹ Goal

Chaining Cognitive Schematics may allow to achieve a form of planning by reasoning, where contexts and goals may typically form a chain of subgoals. For instance the following

InitialContext & A1 ⟹ SubGoal1
SubGoal1 & A2 ⟹ SubGoal2
SubGoal2 & A3 ⟹ FinalGoal

would suggest that one possible way to achieve FinalGoal is to chain actions A1 to A3, provided that InitialContext is true.

There are multiple ways to represent Cognitive Schematics in OpenCog. In an atemporal environment an ImplicationLink (or ExtensionalImplicationLink) may suffice

ImplicationLink <TV>
  AndLink
    Context
    Action
  Goal

In a temporal environment one may need to use PredictiveImplicationLink to capture the delay between actions and effects, such

PredictiveImplicationLink <TV>
  AndLink
    Context
    Action
  Goal

Other ways are possible as well. Managing uncertainty becomes possible by considering the <TV> associated to the implications.

Also, one would typically need to wrap the action in a predicate indicating the willful capacity to trigger the action independently of its actually probability of being triggered (and without triggering it), such as

PredictiveImplicationLink <TV>
  AndLink
    Context
    do(Action)
  Goal

as in Judea Pearl's Do-Calculus. Maybe introducing a DoLink would make sense.

The idea of course is that these Cognitive Schematics can be both learned and utlized by OpenCog. Components using Cognitive Schematics are OpenPsi for planning and URE for inference control.

See also

Glossary: Cognitive Schematic