Beyond Behavior: Proposing an Independent Science of Intention

Behavioral science excels at predicting effects, but prediction alone cannot legitimize action in high-stakes systems. We propose an independent Science of Intention that (i) treats intention as a runtime object (iFlw), (ii) requires proven reversibility, and (iii) demands a live, auditable link between declared purpose and observed behavior. In practice, authorization becomes a single readiness decision that combines three live signals: Wʳ, a readiness measure derived from capability tempered by awareness; iFlwʳ, a traceable flow that binds intent to policy, plan, action, and outcome with a tested rollback path; and μʳ, an indicator that behavior is converging toward the stated intent. We present the runtime framing, a gate policy, audit primitives, three applied vignettes, and limits — positioning intention-centric control as a complement to behavioral prediction, causal modeling, and robust AI safety.