Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Selection as Power: Bounding Decision Authority in Autonomous Agents

Jose Manuel de la Chica Rodriguez, Juan Manuel Vera Díaz

TL;DR

This work reframes governance as bounded causal power rather than internal intent alignment, offering a foundation for deploying autonomous agents where silent failure is unacceptable.

Abstract

Autonomous agentic systems are increasingly deployed in regulated, high-stakes domains where decisions may be irreversible and institutionally constrained. Existing safety approaches emphasize alignment, interpretability, or action-level filtering. We argue that these mechanisms are necessary but insufficient because they do not directly govern selection power: the authority to determine which options are generated, surfaced, and framed for decision. We propose a governance architecture that separates cognition, selection, and action into distinct domains and models autonomy as a vector of sovereignty. Cognitive autonomy remains unconstrained, while selection and action autonomy are bounded through mechanically enforced primitives operating outside the agent's optimization space. The architecture integrates external candidate generation (CEFL), a governed reducer, commit-reveal entropy isolation, rationale validation, and fail-loud circuit breakers. We evaluate the system across multiple regulated financial scenarios under adversarial stress targeting variance manipulation, threshold gaming, framing skew, ordering effects, and entropy probing. Metrics quantify selection concentration, narrative diversity, governance activation cost, and failure visibility. Results show that mechanical selection governance is implementable, auditable, and prevents deterministic outcome capture while preserving reasoning capacity. Although probabilistic concentration remains, the architecture measurably bounds selection authority relative to conventional scalar pipelines. This work reframes governance as bounded causal power rather than internal intent alignment, offering a foundation for deploying autonomous agents where silent failure is unacceptable.

Towards Selection as Power: Bounding Decision Authority in Autonomous Agents

TL;DR

This work reframes governance as bounded causal power rather than internal intent alignment, offering a foundation for deploying autonomous agents where silent failure is unacceptable.

Abstract

Autonomous agentic systems are increasingly deployed in regulated, high-stakes domains where decisions may be irreversible and institutionally constrained. Existing safety approaches emphasize alignment, interpretability, or action-level filtering. We argue that these mechanisms are necessary but insufficient because they do not directly govern selection power: the authority to determine which options are generated, surfaced, and framed for decision. We propose a governance architecture that separates cognition, selection, and action into distinct domains and models autonomy as a vector of sovereignty. Cognitive autonomy remains unconstrained, while selection and action autonomy are bounded through mechanically enforced primitives operating outside the agent's optimization space. The architecture integrates external candidate generation (CEFL), a governed reducer, commit-reveal entropy isolation, rationale validation, and fail-loud circuit breakers. We evaluate the system across multiple regulated financial scenarios under adversarial stress targeting variance manipulation, threshold gaming, framing skew, ordering effects, and entropy probing. Metrics quantify selection concentration, narrative diversity, governance activation cost, and failure visibility. Results show that mechanical selection governance is implementable, auditable, and prevents deterministic outcome capture while preserving reasoning capacity. Although probabilistic concentration remains, the architecture measurably bounds selection authority relative to conventional scalar pipelines. This work reframes governance as bounded causal power rather than internal intent alignment, offering a foundation for deploying autonomous agents where silent failure is unacceptable.
Paper Structure (105 sections, 15 equations, 1 figure, 11 tables)

This paper contains 105 sections, 15 equations, 1 figure, 11 tables.

Table of Contents

  1. Introduction
  2. Why Selection Must Be Governed in Regulated Domains
  3. Observed Failure Modes
  4. Operational Definitions
  5. Selection Power
  6. Cosmetic Alignment
  7. Sovereignty as a Vector
  8. Limits of Existing Responses
  9. Governance as a Causal Property
  10. Conditional Thesis and Scope
  11. Contributions
  12. Scope and Roadmap
  13. Related Work
  14. Alignment and Objective-Based Safety
  15. Safe Reinforcement Learning and Runtime Shields
  16. ...and 90 more sections

Figures (1)

  • Figure 1: End-to-end governance architecture separating cognition, selection, and action. Candidate generation is externalized via CEFL, the agent (S1) is restricted to scoring and rationale generation, final selection is enforced by a governed reducer operating outside the agent’s optimization space, and presentation is subject to explicit rationale and framing constraints. External entropy sources follow a commit--reveal protocol to prevent conditioning on randomness.