Somatic in the East, Psychological in the West?: A Clinically-Grounded Evaluation of Cross-Cultural Depression Symptoms in LLMs

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mental Health & Psychological Support, Social Sciences & Behavioral Studies · Depth: Expert, short

Summary

The paper "Somatic in the East, Psychological in the West?: A Clinically-Grounded Evaluation of Cross-Cultural Depression Symptoms in LLMs" by Sakai, An, Kang, and Kwak, presented at C3NLP 2026, evaluates whether general-purpose LLMs reproduce clinically established cross-cultural patterns in depression symptoms. Clinical psychology notes Western populations emphasize emotional symptoms, while East Asian populations report more somatic symptoms. The researchers used prompts grounded in clinical descriptions, examining LLM responses under different cultural personas and languages. They found LLMs struggle to reproduce expected cultural patterns when prompted in English. Prompting in major Eastern languages improved alignment in some configurations, indicating partial activation of cultural knowledge. However, LLM behavior is largely dominated by a strong, culture-invariant hierarchy of depression symptoms that often overrides cultural cues, revealing limitations for mental health applications.

Key takeaway

For AI Scientists and NLP Engineers developing LLMs for mental health applications, recognize that current models struggle to accurately reflect cross-cultural depression symptom patterns. Your English prompts may yield responses dominated by a culture-invariant symptom hierarchy, overriding cultural cues. Consider incorporating diverse, language-specific training data and evaluating models rigorously across multiple cultural contexts and languages to mitigate these biases and improve clinical relevance.

Key insights

LLMs struggle to reproduce cross-cultural depression symptom patterns, especially in English, due to a dominant culture-invariant symptom hierarchy.

Principles

• Clinical psychology documents cultural differences in depression symptoms.
• LLMs exhibit a strong, culture-invariant symptom hierarchy.
• Language cues can partially activate cultural knowledge in LLMs.

Method

Evaluated general-purpose LLMs using clinically-grounded depression prompts. Examined responses under cultural personas and languages (English, major Eastern languages).

In practice

• LLMs have limitations for mental health applications.
• Prompting in native languages can partially activate cultural knowledge.

Topics

• Large Language Models
• Mental Health Applications
• Cross-Cultural Psychology
• Depression Symptoms
• Prompt Engineering
• Cultural Bias

Best for: AI Scientist, NLP Engineer, Research Scientist

Related on AIssential

• Machine ‘culture’ beyond language — Language reliably shifts an LLM's "psychological profile," influencing social orientation and cognitive style.
• Explainable Detection of Depression Status Shifts from User Digital Traces — A BERT-LLM framework detects mental health shifts from digital traces, providing explainable, temporally coherent reports.
• Linguistic Symptoms: Augmented Generation by Retrieval and Reasoning in LLMs under Portuguese-English Variation in Medical Contexts — Linguistic variation significantly impacts smaller LLMs in medical contexts, with larger models showing greater cross-language robustness.
• AI’s fluency in other languages hides a Western worldview that can mislead users − a scholar of Indonesian society explains — LLMs exhibit "epistemological persistence," retaining a Western worldview despite multilingual fluency due to English-centric training data.
• MADE: Beyond Scoring via a Multilingual Agentic Diagnosing Engine for Fine-Grained Evaluation Insights — MADE offers a structured, agentic approach for fine-grained, multilingual LLM evaluation diagnosis, moving beyond simple benchmark scores.
• LLMs Infer Cultural Context but Fail to Apply It When Responding — LLMs infer cultural context but often fail to apply it in responses without explicit sequential prompting, revealing cultural biases.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.