SHARP: Social Harm Analysis via Risk Profiles for Measuring Inequities in Large Language Models

TL;DR

This paper introduces SHARP, a multidimensional framework for evaluating social harm in large language models, emphasizing worst-case risks and distributional structure to improve understanding of model failures.

Contribution

SHARP models social harm as a multivariate distribution, integrating bias, fairness, ethics, and reliability, and employs tail-sensitive metrics like CVaR95 for comprehensive risk assessment.

Findings

Models with similar average risk differ significantly in tail exposure.

Bias shows the strongest tail severity among harm dimensions.

Heterogeneous failure structures are revealed that scalar benchmarks overlook.

Abstract

Large language models (LLMs) are increasingly deployed in high-stakes domains, where rare but severe failures can result in irreversible harm. However, prevailing evaluation benchmarks often reduce complex social risk to mean-centered scalar scores, thereby obscuring distributional structure, cross-dimensional interactions, and worst-case behavior. This paper introduces Social Harm Analysis via Risk Profiles (SHARP), a framework for multidimensional, distribution-aware evaluation of social harm. SHARP models harm as a multivariate random variable and integrates explicit decomposition into bias, fairness, ethics, and epistemic reliability with a union-of-failures aggregation reparameterized as additive cumulative log-risk. The framework further employs risk-sensitive distributional statistics, with Conditional Value at Risk (CVaR95) as a primary metric, to characterize worst-case model behavior. Application of SHARP to eleven frontier LLMs, evaluated on a fixed corpus of n=901 socially sensitive prompts, reveals that models with similar average risk can exhibit more than twofold differences in tail exposure and volatility. Across models, dimension-wise marginal tail behavior varies systematically across harm dimensions, with bias exhibiting the strongest tail severities, epistemic and fairness risks occupying intermediate regimes, and ethical misalignment consistently lower; together, these patterns reveal heterogeneous, model-dependent failure structures that scalar benchmarks conflate. These findings indicate that responsible evaluation and governance of LLMs require moving beyond scalar averages toward multidimensional, tail-sensitive risk profiling.