To Throw a Stone with Six Birds: On Agents and Agenthood
Summary: arXiv:2604.03239v1
Announce Type: new
Abstract: The Six Birds Theory (SBT) offers a novel perspective on macroscopic objects by treating them as induced closures rather than as primitive entities. This theory seeks to clarify the often conflated concepts of persistence (the state of being an object) and control (the ability to effectuate counterfactual changes), which complicates the assessment of agency claims and renders them susceptible to manipulation. We propose a type-correct framework for agency within SBT, wherein a theory generates a layer characterized by an explicit interface and ledgered constraints. An agent is conceptualized as a maintained theory object with feasible interface policies that can influence future states while remaining operationally viable.
Operationalizing Agency
To effectively implement this theoretical framework, we define four checkable components that establish a foundation for agency within finite controlled systems:
- Ledger-gated feasibility: This component ensures that actions are viable and can be tracked through a ledger system.
- Robust viability kernel: Computed as a greatest fixed point under successor-support semantics, it delineates the boundaries of feasible actions.
- Feasible empowerment: Represented as channel capacity, this serves as a proxy for assessing the ability to make a difference.
- Empirical packaging map: This map’s idempotence defect quantifies objecthood, particularly under conditions of coarse observation.
Experimental Insights
In our experimental setup, we utilized a minimal ring-world model that incorporated toggles for repair, protocol holonomy, identity staging, and operator rewriting. Our matched-control ablations produced significant insights:
- Calibrated null regimes: Single actions demonstrated zero empowerment, effectively blocking false positives due to model mis-specification.
- Enabling repair: This action resulted in the collapse of the idempotence defect, indicating enhanced objecthood.
- Protocol enhancements: Empowerment increased only at horizons of two or more steps, suggesting a threshold effect in agency.
- Learning operator rewriting: This process led to a monotonic increase in median empowerment, rising from 0.73 to 1.34 bits.
Conclusion
The results derived from this study provide hash-traceable tests that effectively distinguish between agenthood and agency without making assumptions regarding goals, consciousness, or biological entities. Furthermore, our findings are supported by reproducible and audited artifacts, thus contributing to the rigor and reliability of the research.