“Automatic Multi-Modal Perception of Students and Classrooms” - Jacob Whitehill
Machine learning offers powerful new ways of analyzing video of school classrooms that are difficult or impractical to implement with human coding. Extracted data can provide fine-grained feedback to teachers, measure the outcome of an educational intervention, or facilitate automated instruction by an AI agent. This talk will present three projects: (1) estimating student engagement with computer vision; (2) exploring the relationship between thermal comfort and learning; and (3) analyzing positive and negative climate, as defined by the Classroom Assessment Scoring System (CLASS), using multi-modal machine learning. Finally, the talk will discuss privacy issues as well as a methodological danger of using automatic classifiers for educational measurement.