Assessing Representation in Minds and Artificial Systems

Representation is a fundamental concept in the study of intelligent systems, and yet it is employed loosely both across and within disciplines. To improve our understanding of cognition, clarifying and expanding this concept is imperative. Moreover, as the capabilities of artificial agents increasingly grow, a need arises to compare these algorithms against humans and other biological systems. Thus, we propose a workshop to improve the understanding of what representations may be implemented in biological and artificial intelligences. We hope not only to engage in discussion about what truly constitutes representation in both systems, but also to propose methodology for identifying, measuring and comparing them. We will consider questions such as: What would constitute a minimal representation that still enables flexible behavior? Are there core components for composing representations? How do we assess whether an artificial or biological agent has a representation, and are there alternate models for understanding agents' interaction with the world? In answering these questions, we hope to bring together speakers from various disciplines and open opportunities for synthesis. This workshop aims to help solidify the shifting grounds of representation and pave the way for future experimental and theoretical work.