Predicting Psychological Well-Being from Spontaneous Speech using LLMs

2026-05-11

AI Summary

This study investigates the feasibility of predicting individual psychological well-being (Ryff PWB) scores from transcribed spontaneous speech—lasting only a few minutes—in an unsupervised, zero-shot setting using large language models (LLMs). Leveraging domain-specific prompts crafted by clinical psychology and linguistics experts, the approach systematically extracts semantic cues through twelve instruction-tuned LLMs, including Llama-3, Mistral, Gemma, and Phi-4. Model interpretability is examined via Spearman correlation analysis, keyword cloud visualization, and statistical bias assessment. The method achieves a peak correlation coefficient of 0.8 on 80% of the data, providing the first empirical evidence that LLMs can effectively capture psychologically meaningful semantic features related to well-being without task-specific training, while also uncovering the linguistic patterns underpinning model predictions and potential sources of bias.

Abstract

We investigate the use of Large Language Models (LLMs) for zero-shot prediction of Ryff Psychological Well-Being (PWB) scores from spontaneous speech. Using a few minutes of voice recordings from 111 participants in the PsyVoiD database, we evaluated 12 instruction-tuned LLMs, including Llama-3 (8B, 70B), Ministral, Mistral, Gemma-2-9B, Gemma-3 (1B, 4B, 27B), Phi-4, DeepSeek (Qwen and Llama), and QwQ-Preview. A domain-informed prompt was developed in collaboration with experts in clinical psychology and linguistics. Results show that LLMs can extract semantically meaningful cues from spontaneous speech, achieving Spearman correlations of up to 0.8 on 80\% of the data. Additionally, to enhance explainability, we conducted statistical analyses to characterise prediction variability and systematic biases, alongside keyword-based word cloud analyses to highlight the linguistic features driving the models' predictions.

Problem

• Psychological Well-Being
• Spontaneous Speech
• Large Language Models
• Zero-shot Prediction
• Ryff PWB

Innovation

• zero-shot prediction
• large language models
• psychological well-being
• spontaneous speech
• explainable AI