MIT Department: Comparative Media Studies
I am a rising senior Computer Science Major at the University of Maryland, Baltimore County. I am driven by a desire to serve my community through STEM outreach. Broadly my research interests lie in the intersection between Computer Science, Social Sciences, and Education, where I can explore applications of computing to improve the quality and accessibility of STEM education. I am also interested in using Computational Cognitive Science in order to understand how we learn and create models that mirror and possibly improve on that process. Aside from my academic and research endeavors, I enjoy basketball, playing video games, listening to jazz, rnb, rap, and ne-soul music (although still not limited to those genres at all), and watching anime/reading manga. I used to play classical piano and trumpet, and I also love cats, but dogs are growing on me.
2019 Research Abstract
Detecting Expressions of Struggle and Confusion in Voice Data
Danilo Symonette1, Garron Hillaire2 and Justin Riech3
Department of Computer Science and Electrical Engineering University of Maryland, Baltimore County
2, 3Department of Comparative Media Studies, Massachusetts Institute of Technology
Effective educators must often engage in difficult conversations surrounding equity, that induce cognitive dissonance, a state of mental discomfort, and require emotional self-regulation (strategies of managing oneself) to achieve successful outcomes. The web application, Teacher Moments¸ seeks to increase teachers’ aptitude for emotion regulation through text-based simulations that mirror these conversations. Monitoring teachers’ emotional experience throughout these simulations could provide insight that can guide intelligent interventions and inform the development of more effective emotional regulation strategies. Previous studies identified struggle and confusion as being related to cognitive dissonance, and others have shown a relationship between cognitive dissonance and the emotional experience. In this study, we explore the use of prosodic (voice) features in training a machine learning model to detect struggle and confusion in 182 voice responses from Teacher Moments. We compare the performance of the models when training with prosodic features, linguistic features, and both. Successful detection will fuel the development of intelligent and personalized interventions that support novice teachers, while also supporting teacher educators with data for coaching.