Re-framing the characteristics of concepts and their relation to learning and cognition in artificial agents