AI-Driven Empathy Assessment Based on Multimodal Physiological Data in Collaboration

Empathy, defined as the ability to understand and share the feelings and needs of others, is crucial in cultivating innovative talents in the age of intelligence. Accurate assessment of empathy is essential in this context. Traditional methods of empathy evaluation mainly rely on self-reported measurements, such as surveys, which are prone to biases and limitations. However, objective methods for quantifying empathy have been relatively underexplored. AI-driven multimodal assessments offer a promising approach to objectively and scientifically study learners' mental states, thus improving the precision and effectiveness of empathy evaluation. Therefore, this study collected data on learners' heart rate, skin conductance, and blood volume pulse during collaborative processes and applied machine learning feature engineering methods and Temporal Graph Networks (TGNs) algorithms to identify empathy levels in the collaborative design process. The results showed that the TGN algorithm achieved a higher accuracy rate of 94%. Furthermore, a correlation analysis between empathy levels and design performance revealed significant differences in project scores among learners with varying levels of empathy (p<0.01). Our research demonstrates that TGNs can serve as a powerful tool for empathy assessment, with broad implications for design education.