Beyond Accuracy: Interpreting Topic Representation in Suicide Ideation Detection Models

2026-06-05

AI Summary

Existing models for detecting suicidal ideation lack a deep understanding of how internal psychological risk factors are represented, relying predominantly on overall accuracy metrics that fall short in high-stakes scenarios demanding both safety and interpretability. This study systematically investigates, for the first time, the impact of topic augmentation on the internal representations of such models. By constructing a topic-augmented dataset and employing representation space visualization alongside geometric feature analysis, the work reveals how key psychosocial risk factors are encoded within the model. The findings demonstrate that topic augmentation substantially enhances the clarity and separability of representation for risk factors associated with vulnerable populations—such as immigration status, family issues, and economic distress—thereby improving both model performance and the interpretability of its internal representational structure.

Abstract

Suicide ideation detection models are typically evaluated using aggregate performance metrics, yet little is known about how they internally represent psychologically meaningful risk factors. In high-stakes mental health applications, understanding these internal representations is essential for safety, transparency, and responsible deployment. In this work, we move beyond accuracy and analyze how suicide detection models trained on original and topic-augmented datasets encode psychological risk factors in their internal representation space. Using visualization and geometric analysis, we examine the coherence and separability of topic-related features. Our results show that topic-aware augmentation increases the clarity and distinctness of underrepresented psychosocial risk factors such as immigration, family issues, and financial crisis. These findings suggest that augmentation not only improves model performance but also leads to more structured and interpretable internal representations.

Problem

• suicide ideation detection
• topic representation
• psychological risk factors
• model interpretability
• mental health AI

Innovation

• topic-aware augmentation
• interpretable representations
• suicide ideation detection
• psychosocial risk factors
• geometric analysis