ACE-Align: Attribute Causal Effect Alignment for Cultural Values under Varying Persona Granularities

TL;DR

ACE-Align introduces a causal-effect framework to align large language models with diverse cultural values by accounting for within-group heterogeneity through demographic attributes, improving cultural sensitivity and geographic equity.

Contribution

It presents a novel causal-effect approach that considers demographic attribute shifts, addressing within-group heterogeneity in cultural value alignment for LLMs.

Findings

Outperforms baselines across 14 countries and various persona granularities.

Reduces geographic alignment gap from 9.81 to 4.92 points.

Achieves largest gains in African regions (+8.48 points).

Abstract

Ensuring that large language models (LLMs) respect diverse cultural values is crucial for social equity. However, existing approaches often treat cultural groups as homogeneous and overlook within-group heterogeneity induced by intersecting demographic attributes, leading to unstable behavior under varying persona granularity. We propose ACE-Align (Attribute Causal Effect Alignment), a causal-effect framework that aligns how specific demographic attributes shift different cultural values, rather than treating each culture as a homogeneous group. We evaluate ACE-Align across 14 countries spanning five continents, with personas specified by subsets of four attributes (gender, education, residence, and marital status) and granularity instantiated by the number of specified attributes. Across all persona granularities, ACE-Align consistently outperforms baselines. Moreover, it improves geographic equity by reducing the average alignment gap between high-resource and low-resource regions from 9.81 to 4.92 points, while Africa shows the largest average gain (+8.48 points). Code is available at https://github.com/Wells-Luo/ACE-Align.