Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Transparent but Powerful: Explainability, Accuracy, and Generalizability in ADHD Detection from Social Media Data

D. Wiechmann, E. Kempa, E. Kerz, Y. Qiao

TL;DR

This paper uses both shallow machine learning models and deep learning approaches, including BiLSTM and transformer-based models, to analyze linguistic patterns in ADHD-related social media text, uncovering key linguistic features associated with ADHD that could contribute to more effective digital screening tools.

Abstract

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent mental health condition affecting both children and adults, yet it remains severely underdiagnosed. Recent advances in artificial intelligence, particularly in Natural Language Processing (NLP) and Machine Learning (ML), offer promising solutions for scalable and non-invasive ADHD screening methods using social media data. This paper presents a comprehensive study on ADHD detection, leveraging both shallow machine learning models and deep learning approaches, including BiLSTM and transformer-based models, to analyze linguistic patterns in ADHD-related social media text. Our results highlight the trade-offs between interpretability and performance across different models, with BiLSTM offering a balance of transparency and accuracy. Additionally, we assess the generalizability of these models using cross-platform data from Reddit and Twitter, uncovering key linguistic features associated with ADHD that could contribute to more effective digital screening tools.

Transparent but Powerful: Explainability, Accuracy, and Generalizability in ADHD Detection from Social Media Data

TL;DR

This paper uses both shallow machine learning models and deep learning approaches, including BiLSTM and transformer-based models, to analyze linguistic patterns in ADHD-related social media text, uncovering key linguistic features associated with ADHD that could contribute to more effective digital screening tools.

Abstract

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent mental health condition affecting both children and adults, yet it remains severely underdiagnosed. Recent advances in artificial intelligence, particularly in Natural Language Processing (NLP) and Machine Learning (ML), offer promising solutions for scalable and non-invasive ADHD screening methods using social media data. This paper presents a comprehensive study on ADHD detection, leveraging both shallow machine learning models and deep learning approaches, including BiLSTM and transformer-based models, to analyze linguistic patterns in ADHD-related social media text. Our results highlight the trade-offs between interpretability and performance across different models, with BiLSTM offering a balance of transparency and accuracy. Additionally, we assess the generalizability of these models using cross-platform data from Reddit and Twitter, uncovering key linguistic features associated with ADHD that could contribute to more effective digital screening tools.

Paper Structure

This paper contains 14 sections, 5 equations, 1 figure, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Experimental Setup
  4. Dataset
  5. A feature framework for ADHD detection models
  6. ADHD Detection Models
  7. Results and Analysis
  8. In-Domain Classification Results
  9. Feature Ablation Results
  10. Generalizability
  11. Conclusion
  12. Ethical Consideration
  13. Limitations
  14. Appendix

Figures (1)

  • Figure 1: Text Feature Dynamics Across Social Media Posts. The figure illustrates the dynamic contours of text features across a series of measurements, focusing on Clauses per Sentence (CS) and Bilogarithmic Type-Token Ratio (bTTR). It visualizes temporal variations in the usage of these features within three consecutive social media posts from (a) an individual diagnosed with ADHD (left) and (b) a control user (right). The z-standardized scores for each sentence depict the dynamic nature of text usage across the posts.