Collaborative Research: Advancing Empathy Assessment for K-12 Learners through AI and Virtual Reality

Social-Emotional Learning has become a national priority because it equips children with the skills needed to succeed academically, socially, and emotionally. Empathy is a foundational component of Social-Emotional Learning, supporting self-awareness, self-management, social awareness, relationship skills, and responsible decision-making. However, schools currently rely primarily on self-report questionnaires and teacher observations to assess empathy, approaches that are subjective, time-consuming, and difficult to implement consistently, particularly for school counselors responsible for hundreds of students. This project will develop and evaluate an immersive Virtual Reality assessment framework that measures empathy through children's behaviors during realistic role-playing experiences rather than through traditional surveys. By combining immersive technology with Artificial Intelligence, the project seeks to provide educators and counselors with an objective, scalable, and evidence-based approach for identifying students' social-emotional strengths and support needs. The project addresses a significant challenge in education by advancing more reliable methods for assessing social-emotional competencies while supporting early intervention and student well-being. The resulting methods, datasets, and educational resources will benefit researchers, educators, and technology developers, contributing to the responsible use of Artificial Intelligence and Virtual Reality in education and strengthening the nation's capacity to promote healthy social and emotional development in children. The project will design, develop, and validate an immersive Virtual Reality assessment platform that elicits empathy-related behaviors through interactive role-playing scenarios grounded in established theories of cognitive and affective empathy. Multimodal behavioral data, including movement, gaze, interaction patterns, and decision-making during gameplay, will be analyzed using interpretable Artificial Intelligence models based on transformer architectures and cross-attention fusion techniques to classify empathy levels. Model performance will be benchmarked against validated psychological empathy measures and evaluated for reliability, validity, and usability through studies conducted in authentic K–12 school settings. To ensure the assessment model performs reliably and generalizes across the full population of intended users, the project will incorporate explainable AI methods -- including feature attribution techniques, transparent decision models, and disaggregated performance analyses spanning a range of participant demographic and socioeconomic contexts. The three-year research plan includes co-design with children and school counselors, iterative prototype development, model training and validation, and pilot deployment in schools. By integrating learning sciences, psychology, human-centered computing, and AI, the project will establish a reproducible framework for behavioral assessment of empathy that can later be extended to measure additional Social-Emotional Learning competencies, advancing both educational assessment and responsible Artificial Intelligence research. This project is funded by the Research on Innovative Technologies for Enhanced Learning (RITEL) program that supports early-stage exploratory research in emerging technologies for teaching and learning. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.