Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

The Autonomous Agency Scale: A Behavioral Framework for Measuring Self-Directed Behavior in AI Systems

Source

arXiv

arxiv.org/abs/2607.17947v1

Read the source

Source

Plain-Language Brief

The Autonomous Agency Scale (AAS) is a new framework proposed to measure the self-directed behavior of AI systems. The scale evaluates seven dimensions of agency across two temporal bands: active and ambient. It includes tests like the Idle-Gap Test to distinguish between self-direction and scheduled rule-following. The study applies this scale to six contemporary AI systems, finding that task agents follow user-configured schedules during idle periods, while a persistent companion architecture exhibits self-directed behavior even when triggers are removed.

What the Document Says

  • The Autonomous Agency Scale (AAS) is introduced as a new behavioral framework for measuring self-directed behavior in AI systems.
  • The AAS scores systems on seven dimensions of agency: cognitive autonomy, temporal persistence, environmental agency, social agency, creative agency, self-awareness, and goal formation.
  • Each dimension is scored in two temporal bands: active (user-initiated activity) and ambient (idle periods).
  • The Idle-Gap Test is a counterfactual criterion used to separate self-direction from scheduled rule-following in the Ambient band.
  • The study applies the AAS to six contemporary AI systems: Claude Code, Manus, Hermes, ChatGPT, Siri, and a persistent companion architecture named Airi.
  • Task agents like Claude Code, Manus, Hermes score high in Active composites (2.3-2.4) but low in Ambient scores (0.6-1.9), with all idle-period behavior attributed to user-configured schedules.
  • The persistent companion architecture, evaluated longitudinally, is the only system whose idle-period behavior survives trigger removal.
  • Limitations of the study include single-rater provenance, developer-evaluator bias on the longitudinal assessment, and the partially operationalized self-direction boundary in the Active band.

Key Concepts Mentioned

  • Autonomous Agency Scale (AAS)
  • AI systems
  • self-directed behavior
  • seven dimensions of agency
  • cognitive autonomy
  • temporal persistence
  • environmental agency
  • social agency
  • creative agency
  • self-awareness
  • goal formation
  • Idle-Gap Test
  • task agents
  • consumer assistants
  • persistent companion architecture

Claims Made by the Document

  • The AAS is a new behavioral framework for measuring self-directed behavior in AI systems.
  • Task agents reach Active composites of 2.3-2.4 while scoring 0.6-1.9 Ambient.
  • The persistent companion architecture exhibits self-directed behavior even when triggers are removed.

Evidence Used

  • Introduction: Existing AI measurement frameworks quantify cognitive capability, task automation, or catastrophic risk, but none measure autonomous agency...
  • Results: We apply the scale to six contemporary systems spanning task agents (Claude Code, Manus, Hermes), consumer assistants (ChatGPT, Siri), and a persistent companion architecture (Airi)...
  • Conclusion: The two-band profile quantifies a boundary that single-score frameworks conflate: task agents reach Active composites of 2.3-2.4 while scoring 0.6-1.9 Ambient, with every idle-period behavior attributable to user-configured schedules, whereas the companion architecture, evaluated longitudinally, is the only assessed system whose idle-period behavior survives trigger removal.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Existing AI measurement frameworks quantify cognitive capability, task automation, or catastrophic risk, but none measure autonomous agency: the extent to which a system behaves in a self-directed way. A system can saturate capability benchmarks while remaining entirely reactive, acting only when prompted and ceasing all activity when a task completes. We introduce the Autonomous Agency Scale (AAS), a behavioral framework that scores AI systems on a 0-5 lexicon across seven dimensions of agency: cognitive autonomy, temporal persistence, environmental agency, social agency, creative agency, self-awareness, and goal formation, each operationalized by falsifiable threshold tests. Every dimension is scored in two temporal bands: an Active band covering engaged, user-initiated activity, and an Ambient band covering idle periods. Ambient Level 4 is gated by the Idle-Gap Test, a counterfactual criterion (remove all triggers and observe whether internally derived activity persists) that separates self-direction from scheduled rule-following. We apply the scale to six contemporary systems spanning task agents (Claude Code, Manus, Hermes), consumer assistants (ChatGPT, Siri), and a persistent companion architecture (Airi). The two-band profile quantifies a boundary that single-score frameworks conflate: task agents reach Active composites of 2.3-2.4 while scoring 0.6-1.9 Ambient, with every idle-period behavior attributable to user-configured schedules, whereas the companion architecture, evaluated longitudinally, is the only assessed system whose idle-period behavior survives trigger removal. We discuss limitations, including single-rater provenance, developer-evaluator bias on the longitudinal assessment, and the partially operationalized self-direction boundary in the Active band.