Extracting the Affective Content of Fidgeting in Adults with ADHD via Machine Learning and a Hand-held Soft Tangible Device

Abstract

This paper examines the task of extracting affect from the fidgeting behavior of adults with Attention Deficit/Hyperactivity Disorder (ADHD) engaged in emotional self-regulation experiments. We describe a fidgeting appliance developed to capture a tactile signal from spontaneous hand fidgeting, clarify its affordances, and present machine learning methods for processing this data. We show that hand-fidgeting carries a strong affective signal by recognizing the affective state produced by emotion-inducing videos and predicting self-assessed measures of anxiety and anxiety management collected after a standardized stress test. We conclude by discussing the potential impact of practical methods for data collection and analysis of fidgeting behavior on ADHD detection and management.