A Jev-like Decision Model for Fast Collaboration Grounded in Behavior Trees

Fast collaboration requires making quick, good-enough decisions about what to do and who does it, without waiting for slow deliberation. This thesis designs and evaluates a lightweight, “Jev-like” decision model that makes such collaboration decisions directly from an actionable task representation, a behavior tree: deciding which task to pursue and who should perform it (agent, partner, or both) as a fast, grounded function of the tree and the current state. The goal is a decision rule that is fast, interpretable, and easy to extend.