When Rating Scales Fall Short: LLM-Assisted Discovery of ADHD Signals in Turkish Teacher Narratives

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Health & Wellbeing — Artificial Intelligence & Machine Learning, Mental Health & Psychological Support, Health & Medical Research · Depth: Expert, quick

Summary

A study by Karacan, Songur, Ozaslan, and Iseri, presented at CLPsych 2026, explores the potential of large language model (LLM)-assisted natural language processing (NLP) to uncover Attention Deficit Hyperactivity Disorder (ADHD) signals in open-ended Turkish teacher narratives. Analyzing de-identified teacher evaluation forms, which included both Conners' Teacher Rating Scale–Revised Short Form (CTRS-R:S) scores and narrative text, the researchers compared predictive signals from both data types. They found that structured assessments sometimes failed to differentiate ADHD from non-ADHD students, while narrative-based models successfully captured distinct behavioral patterns. These narrative signals showed minimal overlap with cases missed by the narrative model itself, indicating complementary information. An LLM-assisted theme discovery pipeline further revealed distinct attention, behavioral, and family-related patterns, underscoring NLP's capability to enhance traditional ADHD screening tools.

Key takeaway

For clinical psychologists and NLP engineers developing diagnostic tools for neurodevelopmental disorders, this research suggests integrating LLM-assisted natural language processing into your assessment workflows. You should consider analyzing open-ended teacher narratives alongside standardized rating scales to capture crucial, complementary behavioral and family-related signals that improve diagnostic accuracy. This approach can enhance the comprehensiveness of ADHD screening, revealing patterns overlooked by traditional structured instruments.

Key insights

LLM-assisted NLP can extract complementary ADHD signals from teacher narratives missed by structured rating scales.

Principles

• Teacher narratives encode unique ADHD signals.
• Structured and narrative data offer complementary insights.
• LLMs can discover clinical themes in unstructured text.

Method

The study applied an LLM-assisted theme discovery pipeline to de-identified Turkish teacher narratives, comparing predictive signals with structured CTRS-R:S scores to identify distinct behavioral patterns.

In practice

• Integrate NLP for richer ADHD assessment data.
• Use LLMs to analyze qualitative clinical notes.
• Combine structured and unstructured data for diagnosis.

Topics

• ADHD Diagnosis
• Natural Language Processing
• Large Language Models
• Clinical Psychology
• Teacher Narratives
• Neurodevelopmental Disorders

Best for: AI Scientist, Research Scientist, NLP Engineer

Related on AIssential

• Reading between the Lines: Leveraging Large Language Models for Global Dementia and Depression Assessment from Clinical Interviews — LLMs can effectively assess neuropsychiatric disorder severity from speech, enabling automated screening.
• Stringology-Based Motif Discovery from EEG Signals: an ADHD Case Study — Stringology-based motif discovery offers a novel, complementary approach to analyze EEG dynamics in neurodevelopmental disorders like ADHD.
• A domain-adapted large language model to support clinicians in psychiatric clinical practice — Domain-adapted LLMs can significantly enhance psychiatric clinical workflows, improving accuracy and efficiency.
• Beyond Scalar Scores: Exploring LLM-based Metrics for Clinical Significance Evaluation in Radiology Reports — LLM-based metrics for radiology reports show discrimination bias, over-penalizing harmless variations while detecting errors.
• Predicting Post-Traumatic Epilepsy from Clinical Records using Large Language Model Embeddings — LLM embeddings from clinical records enhance early prediction of Post-Traumatic Epilepsy when fused with tabular data.
• T2D-Bench: Evidence-Gated Evaluation of LLM Outputs for Type 2 Diabetes Using a Multi-Layer Clinical-Lifestyle Knowledge Graph — LLMs require evidence-gated evaluation against clinical knowledge graphs to ensure medical accuracy and guideline compliance.

Open in AIssential →

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