When PCOS Meets Eating Disorders: An Explainable AI Approach to Detecting the Hidden Triple Burden

2026-04-15 · Source: Computation and Language · Field: Health & Wellbeing — Artificial Intelligence & Machine Learning, Medical Specialties & Subspecialties, Mental Health & Psychological Support · Depth: Expert, quick

Summary

Researchers developed small, open-source language models to detect the "triple burden" of polycystic ovary syndrome (PCOS), body image distress, and disordered eating in social media posts. The project aimed to provide transparent detection of these co-occurring conditions, which are often missed by traditional natural language processing methods. A dataset of 1,000 PCOS-related posts was collected from six subreddits and manually annotated by two experts using the Lee et al. (2017) clinical framework. Three models—Gemma-2-2B, Qwen3-1.7B, and DeepSeek-R1-Distill-Qwen-1.5B—were fine-tuned using Low-Rank Adaptation to generate structured explanations with textual evidence. The top-performing model achieved 75.3% exact match accuracy on 150 held-out posts, demonstrating robust comorbidity detection and strong explainability, though performance decreased with increased diagnostic complexity.

Key takeaway

For public health researchers and NLP engineers developing mental health screening tools, this work suggests that small, explainable language models can effectively identify complex, co-occurring conditions like PCOS and eating disorders from social media data. Your focus should be on leveraging these models for early screening and risk identification, rather than autonomous diagnosis, due to performance declines with diagnostic complexity. Consider integrating similar fine-tuned models to enhance transparency and provide textual evidence for detected conditions.

Key insights

Explainable AI can detect co-occurring health conditions like PCOS and eating disorders from social media data.

Principles

• Small LMs can achieve robust comorbidity detection.
• Explainability enhances diagnostic transparency.

Method

Fine-tuning small language models (Gemma-2-2B, Qwen3-1.7B, DeepSeek-R1-Distill-Qwen-1.5B) with Low-Rank Adaptation on annotated social media data to generate structured explanations.

In practice

• Screen social media for early detection of health comorbidities.
• Use LMs for grounded explainability in health applications.

Topics

• PCOS
• Eating Disorders
• Explainable AI
• Natural Language Processing
• Social Media Analysis

Best for: AI Scientist, Research Scientist, NLP Engineer

Related on AIssential

• Beyond the Blood Draw: Explainable Machine Learning for Non-Invasive Dysglycemia Risk Screening — Explainable ML models can non-invasively screen for dysglycemia risk using readily available data, outperforming existing clinical scores.
• AI scans 400,000 Reddit posts and finds hidden Ozempic side effects — AI-driven social media analysis can rapidly identify underreported drug side effects missed by traditional clinical trials.
• Speech as a biomarker for supported diagnosis of major depressive disorder using self-supervised representations — Self-supervised speech representations offer robust, objective biomarkers for MDD diagnosis.
• Symb-xMIL: Symbolic Explanations for Multiple Instance Learning in Digital Pathology — Symb-xMIL explains MIL model decisions by aligning their behavior with human-readable logical rules over semantic features.
• Towards a General Intelligence and Interface for Wearable Health Data — Large-scale pretraining on wearable data creates a foundation model for diverse health insights.
• Exploring the feasibility of conversational diagnostic AI in a real-world clinical study — Supervised conversational AI for pre-visit history taking is feasible, safe, and well-received by patients and clinicians.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.