You Know What I'm Saying: Designing Conversational Strategies of AI Agent for Tip of the Tongue Phenomenon

Abstract

The phenomenon of being unable to recall a specific word or term from memory is called "tip-of-the-tongue"(TOT). Despite its significance, potential applications for solving TOT from the user experience (UX) and human-computer interface (HCI) perspectives are yet to be fully explored. With the advent of advanced artificial intelligence (AI) interfaces and technologies, new opportunities have emerged to address TOT. This study aims to investigate the applicability of AI in addressing TOT and the roles of conversation design strategies in mitigating this phenomenon. To this end, we developed a prototype called "Tip-pong,"which is a turn-taking user interface that provides semantic and episodic clues, as well as targeted and optional answers to assist users in resolving TOT. Through a quantitative and qualitative study involving 23 participants, we evaluated the prototype and found that while the type of clue had no significant impact on usability, users preferred optional answers over targeted answers. Additionally, we found that allowing users to experience the process of approaching the answer on their own was more effective than providing the correct answer immediately. Based on these findings, we discuss the design implications for the development of AI agents to solve TOT in everyday life.