Functional Intentionality Test: Measuring AI Goal-Directed Behavior

A new arXiv preprint defines intentionality for AI systems as a behavioral profile—not consciousness—based on purpose, foresight, volition, temporal commitment, and coherence. The authors argue these properties are design-contingent, shaped by architectural choices like memory persistence and planning depth. They introduce the Functional Intentionality Test (FIT), a multidimensional framework to quantify intentional-like behavior across five observable dimensions. The paper positions this as a tool for governance and accountability, enabling detection of autonomous, goal-directed behavior in AI.