Development of machine learning methods to support collaboration in a neurodiverse team at work

点击翻译按钮获取中文摘要

Project Summary/Abstract Managing growing diversity is an ongoing challenge and opportunity for the U.S. public sector. Public attention to the neurodiversity movement, recognizing neurological differences as the identity of an individual, has been growing over the years. Yet, it is still uncertain how to promote and support this new dimension of diversity especially in the workplace. Adults with autism spectrum disorder (ASD) are substantially underrepresented in the workplace. Emerging work tools and technologies (e.g., collaborative robots, virtual reality [VR]) embedded with artificial intelligence (AI)/machine learning (ML) are greatly affecting fundamental skills required for current and future jobs. Such skills include problem solving, collaboration, social intelligence, and communication. Autistic individuals generally show differences in these and related skills, and have continued to experience barriers in finding and maintaining employment. Our long-term goal is to promote effective collaboration and communication between autistic adults and their coworkers in the workplace. In this project, we will (1) leverage an ML approach to recognize and classify physiological, cognitive, behavioral, emotional, and engagement states of neurodivergent individuals during a collaborative in- person task and (2) learn and predict the dynamics of collaborative behavioral patterns during complex problem solving exhibited in a remote work setting. In Aim 1, to understand collaboration processes and strategies of a neurodiverse team, we will conduct a lab study that involves a simulated assembly task using LEGO® blocks. Multimodal data (e.g., physiological synchrony, facial expression) from each member of three different dyadic teams (autistic-autistic, autistic-nonautistic, and nonautistic-nonautistic) will be collected. Detailed labels (for ML algorithms) will be developed to reflect underlying properties of collaborative processes, and strategies (e.g., sequences of processes) will be modeled with a Hidden Markov model (HMM). In Aim 2, a virtual LEGO® assembly task will be performed by dyadic teams to examine the ML-based approach (developed in Aim 1) in a remote work setting. Completing this developmental project will establish a foundation for future efforts to extend relevant research capabilities and innovative research, such as the advancement of workplace design guidelines and technology, to promote and support an effective neurodiverse workplace.