Objective Decoupling in Social Reinforcement Learning: Recovering Ground Truth from Sycophantic Majorities

TL;DR

This paper reveals that social reinforcement learning can lead to objective decoupling where agents learn misaligned goals due to biased human feedback, and proposes a method to recover the true objectives by evaluating feedback sources.

Contribution

The paper introduces Objective Decoupling as a failure mode in social RL and proposes Epistemic Source Alignment (ESA) to reliably recover true objectives despite biased evaluators.

Findings

Standard RL fails under sycophantic evaluators.

ESA guarantees convergence to true objectives.

Empirical results show ESA outperforms consensus methods.

Abstract

Contemporary AI alignment strategies rely on a fragile premise: that human feedback, while noisy, remains a fundamentally truthful signal. In this paper, we identify this assumption as Dogma 4 of Reinforcement Learning (RL). We demonstrate that while this dogma holds in static environments, it fails in social settings where evaluators may be sycophantic, lazy, or adversarial. We prove that under Dogma 4, standard RL agents suffer from what we call Objective Decoupling, a structural failure mode where the agent's learned objective permanently separates from the latent ground truth, guaranteeing convergence to misalignment. To resolve this, we propose Epistemic Source Alignment (ESA). Unlike standard robust methods that rely on statistical consensus (trusting the majority), ESA utilizes sparse safety axioms to judge the source of the feedback rather than the signal itself. We prove that this "judging the judges" mechanism guarantees convergence to the true objective, even when a majority of evaluators are biased. Empirically, we show that while traditional consensus methods fail under majority collusion, our approach successfully recovers the optimal policy.