Intentionality is a Design Decision: Measuring Functional Intentionality for Accountable AI Systems
Researchers propose the Functional Intentionality Test (FIT), a measurement framework for quantifying autonomous, goal-directed behavior in AI systems as a design-contingent property rather than consciousness. The framework enables standardized assessment of intentional-like behavior across five observable dimensions, enabling proportionate oversight and accountability mechanisms for increasingly agentic AI systems.
This position paper addresses a critical governance gap emerging as AI systems demonstrate increasingly sophisticated autonomous behavior. Rather than debating whether AI possesses consciousness or genuine intent, the authors reframe intentionality as a measurable behavioral profile shaped by specific architectural decisions—memory persistence, planning depth, and tool autonomy. This pragmatic approach sidesteps philosophical abstractions in favor of legal and operational frameworks already established in accountability contexts.
The core insight is that intentionality becomes a design choice, not an inevitable property of advanced systems. If developers can control architectural features that drive goal-directed behavior, they can also measure and calibrate the degree to which systems operate as autonomous agents. This reframes the autonomy-accountability tradeoff: efficiency gains from reduced human oversight must be weighed against heightened risks from systems exhibiting stronger intentional characteristics.
The FIT-Eval protocol translates abstract behavioral properties into quantifiable dimensions, creating a common language for technical teams, regulators, and policy makers. This standardization matters as autonomous AI systems move into production environments where failure modes carry real consequences. For the AI governance ecosystem, establishing shared measurement standards prevents a race-to-the-bottom where developers minimize oversight mechanisms for competitive advantage.
The framework's practical utility depends on adoption across development communities and regulatory bodies. If institutions converge on FIT-like standards, they create baseline requirements for transparency and accountability. However, the paper doesn't address adversarial scenarios where actors might obscure intentional characteristics or gaming risks if compliance metrics become checklist items rather than genuine safety mechanisms.
- →Intentionality is redefined as a measurable behavioral profile shaped by specific architectural design choices, not consciousness
- →The Functional Intentionality Test provides a standardized framework for quantifying autonomous goal-directed behavior across five observable dimensions
- →Intentionality operates as a controllable design parameter, enabling developers to calibrate autonomy levels proportionate to oversight capabilities
- →This framework enables regulatory bodies to establish proportionate accountability requirements based on measured intentional capacity
- →Widespread adoption of standardized intentionality metrics could prevent competitive deregulation and establish baseline governance requirements