SUMMARY

This panel session will bring together experts in natural language processing, cognitive science, psycholinguistics, socially-intelligent technologies and human-robot interaction to discuss the need for human-oriented research to inform the design and evaluation of future models of dialogue that are adaptive, audience-aware, inclusive, and facilitate naturalistic interaction.  The panellists will discuss recent developments relevant to conversational systems and share examples of how these systems can be improved through increased understanding of human behaviour.

The discussion will touch on several key areas, including the extent to which conversational AI should adapt to humans/users, the importance of explainability, trustworthiness and safety of such systems, which human-like qualities are necessary for effective naturalistic interaction, as well as how to evaluate and improve current state-of-the-art systems. Overall, the panel will explore how a more interdisciplinary approach can inform future directions of conversational systems and aims to provoke discussion about the future of conversational AI.