Modeling Interactive Intelligences

This paper explores modeling the adaptive mechanisms of autonomous agents. The focus is on reflexive interaction as a looping action that creates a “self.” Imitation and play are adaptive components of reflexive interactions that can provide gradual modifications of the agent’s self. Little by little, the autonomous agent reinforces what seems to work and phases out other options through an internal editing process. A performance metrics for reflexive interaction would have to depend on the task at hand within a context, rather than try to be an absolute measure. Evaluation of performance has to be flexible enough to account for multiple intelligences especially when innovation is possible.