Trusted AI and Hospitable Search Systems

Abstract

We aspire to build trusted systems. Near future interactive information retrieval (IIR) systems will feature artificially intelligent interactions with the user. These systems will include dialog, conversational, and hybrids with conventional, but intelligent, IIR systems. Users will inevitably attribute agency to such AI instances. This causes new dimensions of system performance, interaction capacities, and UX. A vision for future systems must consider the development and maintenance of trust. Product and system design will cope with an expanded scope of 'trust surfaces' for user relationships. This talk introduces one dimension for developing trust in an AI search system, organized around the idea of a system capacity for 'hospitable AI'. Future work will focus on other capacities that affect trust development, including task understanding and subtask responsibility negotiation.

Keywords: user-centered search, search process, AI, information system design, search interaction