Overview Index

Reward distribution routine

Reward distribution routine prioritizes sub goals by updating their desirability attributes.

Reward distribution routine distributes reward value among the concepts responsible for the goal satisfaction.


One of super goals is (partially) satisfied. The super goal calculates reward and passes it to reward distribution routine. Reward distribution routine distributes reward among cause concepts responsible for the super goal satisfaction.


Input parameters

1) Reward value.

2) Specified super goal.

Implementation of reward-distribution routine

Every super goal is represented by at least one concept in the main memory.

This “super goal’s deputy concept” has: ConceptId = SuperGoalConceptId

Step #1:

Find cause concepts responsible for the super goal satisfaction

That means to select all records from cause-effect relation table with EffectConceptID = SuperGoalConceptId:

FROM CauseEffectRelation
WHERE EffectConceptID = SuperGoalConceptId

Step #2:

Distribute reward among selected concepts proportionally Coherence attribute.

That means to update desirability attribute of every selected concept.


How does strong AI know what caused super goal satisfaction?

Strong AI knows about cause and effect of everything (including causes for super goal satisfaction) from learning, in particular from experiment.

See also:

Satisfaction level diagram