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

2026-01-19

AI Summary

This work addresses the tendency of existing large language models to treat cultural groups as homogeneous, thereby overlooking internal heterogeneity arising from intersecting demographic attributes—such as gender, education, residence, and marital status—which leads to inconsistent cross-persona behaviors. To mitigate this, we propose the ACE-Align framework, which introduces, for the first time, an attribute causal effect alignment mechanism. By leveraging causal inference, ACE-Align explicitly models the differential impacts of multiple attributes on cultural values, enabling fine-grained alignment. Experiments across 14 countries demonstrate that our method consistently outperforms baselines at all persona granularities, reducing the average alignment gap between high- and low-resource regions from 9.81 to 4.92 and achieving an average improvement of +8.48 points in African regions, thereby significantly enhancing model stability and fairness.

Technology Category

• Humans and AI: Learning Human Values and Preferences
• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• cultural values
• within-group heterogeneity
• persona granularity
• demographic attributes
• large language models

Innovation

• causal effect
• cultural alignment
• persona granularity
• demographic attributes
• LLM fairness