Social robots and language technologies: Insights and open challenges in aligning representations in social interactions

This talk will explore representational alignment in interacting systems in social robot studies and language technologies. The first part will present robots such as the child-friendly Opie robot in a study with children teaching colour names to the robot which had a distinctly different colour space; robot-robot language creation; and the rat-sized iRat robot, modelling SLAM in rodent navigation systems which unexpectedly led to prosocial tasks with rats. The second part will present a computer scientist’s take on how computational representations relate to classical data structures and algorithms; how they differ from biological representations; compare embeddings in artificial NNs with human memory. The talk will finish with open challenges from human cognition, including active construction of task-based representations in analogies and social interactions.