Toward Culturally Aligned LLMs through Ontology-Guided Multi-Agent Reasoning

TL;DR

This paper introduces OG-MAR, a framework that enhances culturally aligned decision making in LLMs by using an ontology-guided multi-agent reasoning approach that incorporates structured cultural values and demographic information.

Contribution

It presents a novel ontology-guided multi-agent reasoning framework that improves cultural alignment and transparency in LLM outputs by integrating structured cultural values and demographic data.

Findings

OG-MAR improves cultural alignment over baselines.

OG-MAR enhances robustness in diverse social survey benchmarks.

The framework produces more transparent reasoning traces.

Abstract

Large Language Models (LLMs) increasingly support culturally sensitive decision making, yet often exhibit misalignment due to skewed pretraining data and the absence of structured value representations. Existing methods can steer outputs, but often lack demographic grounding and treat values as independent, unstructured signals, reducing consistency and interpretability. We propose OG-MAR, an Ontology-Guided Multi-Agent Reasoning framework. OG-MAR summarizes respondent-specific values from the World Values Survey (WVS) and constructs a global cultural ontology by eliciting relations over a fixed taxonomy via competency questions. At inference time, it retrieves ontology-consistent relations and demographically similar profiles to instantiate multiple value-persona agents, whose outputs are synthesized by a judgment agent that enforces ontology consistency and demographic proximity. Experiments on regional social-survey benchmarks across four LLM backbones show that OG-MAR improves cultural alignment and robustness over competitive baselines, while producing more transparent reasoning traces.