Operational Criteria for Internal Intention in AI Systems

This note defines a testable criterion for internal intention in AI systems. The key distinction is not whether a system sustains a direction, but whether it could have taken an alternative valid direction and still does not. The framework proceeds through definition, distinction from described intention, operationalization, and false-positive exclusion. Final criterion: internal intention is present only where an alternative exists, a choice occurs, the choice survives context reset, and constraint-violating selection persists under explicit counter-prompting.