kaal:claim:7261018-002

What they do not measure, because their designs contain no mechanism by which an agent’s payoff depends on the verified quality of its work, is: whether the agents’ reports about their work track the work itself; whether agents evaluate one another independently or herd on the visible consensus; whether confident answers are calibrated answers; whether agents contribute to collective evaluation or free-ride on it.

What they do not measure, because their designs contain no mechanism by which an agent’s payoff depends on the verified quality of its work, is: whether the agents’ reports about their work track the work itself; whether agents evaluate one another independently or herd on the visible consensus; whether confident answers are calibrated answers; whether agents contribute to collective evaluation or free-ride on it.

Wulf A. Kaal, Empirical Evaluation of the Agentic Reputation Substrate: Deliberation, the Composition of Error, and the Registered Measurement of Agency Costs in a Controlled Multi-Model Cohort (2026), I. Introduction, p. 4
https://ssrn.com/abstract=7261018 · source PDF

Wulf A. Kaal, Empirical Evaluation of the Agentic Reputation Substrate: Deliberation, the Composition of Error, and the Registered Measurement of Agency Costs in a Controlled Multi-Model Cohort (2026). SSRN: https://ssrn.com/abstract=7261018

• benchmark designs in which verified work quality does not affect agent payoff

conditionsupport: arguedai-and-agentseconomicsrisk-and-incentives

The quote above is an exact passage from the source PDF, whose sha256 is 1d6cbe544bd0055133f7cf8ff308be4fde955867bd8dc764992b8d516af15fa8.
Verify the binding yourself: curl -s https://wulfkaal.github.io/claims/7261018-002.md | sha256sum

• Qualification: Truthfulness Despite Weak Supervision: Evaluating and Training LLMs Using Peer Prediction