Subjective Logic

Subjective logic is a framework for representing and manipulating uncertain information, particularly opinions and beliefs, by modeling them as probability distributions. Current research focuses on applying subjective logic to improve uncertainty quantification in deep learning models, particularly for tasks involving ambiguous or composite classifications, often using Dirichlet distributions and variations of evidential deep learning. This work aims to enhance the reliability and transparency of AI systems by explicitly representing and managing uncertainty, with applications ranging from autonomous driving to social network analysis and product recommendation systems.