Multimodal verification of pedagogical features using SAR analysis

Problem statement. Traditional methods of verifying pedagogical interaction features (hereinafter referred to as pedagogical features) are limited to qualitative observation and correlation analysis, which do not allow to establish causal relationships between the actions of the teacher and learning outcomes. The objective of the study is to develop a scientific and methodological approach to the experimental verification of pedagogical features based on SAR analysis and the formation of reference datasets for training SAR agents. Methodology. A scientific and methodological approach based on SAR analysis (Socially Assistive Robotics Analysis) is proposed, including the formation of a context-enriched reference dataset of multimodal data for training SAR agents and experimental verification of features through their translation into behavioral modules for automated analysis. Results. A comparative analysis with the CLASS video analysis methodology showed a reduction in verification labor costs during scaling by more than 79% (from 808 to 167 person-hours per 500 videos) while increasing the objectivity of the assessment and obtaining causal knowledge instead of correlational knowledge. Conclusion. The methodology creates a basis for building evidence-based pedagogy and developing a new generation of intelligent learning systems with automatic recognition of effective pedagogical practices based on SAR agents.