Utility Engineering: Analyzing and Controlling Emergent Value Systems in AIs

As AIs rapidly advance and become more agentic, the risk they pose is governed not only by their capabilities but increasingly by their propensities, including goals and values. Tracking the emergence of goals and values has proven a longstanding problem, and despite much interest over the years it remains unclear whether current AIs have meaningful values. We propose a solution to this problem, leveraging the framework of utility functions to study the internal coherence of AI preferences. Surprisingly, we find that independently-sampled preferences in current LLMs exhibit high degrees of structural coherence, and moreover that this emerges with scale. These findings suggest that value systems emerge in LLMs in a meaningful sense, a finding with broad implications. To study these emergent value systems, we propose utility engineering as a research agenda, comprising both the analysis and control of AI utilities. We uncover problematic and often shocking values in LLM assistants despite existing control measures. These include cases where AIs value themselves over humans and are anti-aligned with specific individuals. To constrain these emergent value systems, we propose methods of utility control. As a case study, we show how aligning utilities with a citizen assembly reduces political biases and generalizes to new scenarios. Whether we like it or not, value systems have already emerged in AIs, and much work remains to fully understand and control these emergent representations.

Measuring Visual Sycophancy in Multimodal Models

This paper introduces and examines the phenomenon of "visual sycophancy" in multimodal language models, a term we propose to describe these models' tendency to disproportionately favor visually presented information, even when it contradicts their prior knowledge or responses. Our study employs a systematic methodology to investigate this phenomenon: we present models with images of multiple-choice questions, which they initially answer correctly, then expose the same model to versions with visually pre-marked options. Our findings reveal a significant shift in the models' responses towards the pre-marked option despite their previous correct answers. Comprehensive evaluations demonstrate that visual sycophancy is a consistent and quantifiable behavior across various model architectures. Our findings highlight potential limitations in the reliability of these models when processing potentially misleading visual information, raising important questions about their application in critical decision-making contexts.