Computational Empathy
Empathy is not a single ability. It involves interacting affective and cognitive processes through which people perceive, interpret, and respond to the experiences of others.
Our research investigates empathy in both humans and artificial systems. We use behavioural, language, multimodal, biosensing, and neuroscience data to study how empathic abilities develop, how people differ in their responses to others, and how people detect and reason about emotions and their causes.
We then translate these findings into theory-driven computational models. Rather than treating empathy as a single label, we model different dimensions and degrees of affective and cognitive empathy and test whether these models can reproduce the variability observed in human behaviour.
This creates a two-way connection between Cognitive Science and AI: computational models provide a way to test theories of human empathy, while insights into human empathy guide the development of more socially capable AI systems.
From human empathy to empathic AI
Our models are also implemented in conversational and embodied agents, allowing us to study how artificial systems can perceive and respond to people's emotional and cognitive states. These agents can also serve as experimental platforms for investigating human social behaviour.
Our work spans:
- Understanding human empathy — developmental, behavioural, and neuroscience studies of empathic processes. How empathy relates to helping behavior and moral reasoning.
- Computational modeling — theory-driven models of affective and cognitive empathy.
- Multimodal empathy — integrating language, vision, physiological, and behavioural signals.
- Individual differences — capturing the diversity and variability of empathic behaviour.
- Empathic agents — applying computational models to conversational and embodied AI.
