GENERATING HUMAN-CENTERED EXPLANATIONS FOR SYSTEMS CAPABLE OF FACIAL EXPRESSION RECOGNITION

This thesis studies how to explain facial expression recognition systems to users. It shows that explanations improve users’ understanding of which expressions such systems will predict and why, empowering users to adopt these systems to their advantage, while having an appropriate level of trust and reliance towards them. This thesis develops a novel, human-centered, explanation method based on facial muscle movements and demonstrates its benefits in improving user understanding and engendering appropriate trust compared to existing methods. The thesis also shows that this novel explanation method can be used to personalize facial expression recognition systems to better align with specific users.