A Pilot Study for Understanding Users' Attitudes Towards a Conversational Agent for News Recommendation

Conversational recommender agents have been rapidly developed and applied in various domains (e.g., amusement, e-commerce, tourism) in recent years, to allow users to easily access information or service through natural communication with the system. However, little attention has been paid to the news domain, though some news organizations (e.g., ABC, BBC) have started to deploy news chatbots to engage with audiences. In this work, we performed a pilot study in form of a semi-structured interview for the purpose of knowing important features of recommendations users expect when they interact with a news conversational agent. In particular, in order to acquire users' thoughtful feedback, we implemented a prototype system based on a taxonomy that covers all of the major recommendation-seeking and information-searching goals according to related literature. The interview results reveal users' opinions on various aspects of a conversational agent for news recommendation, including the condition under which they may request/accept the news recommendation by a conversational agent, important features of the conversational news recommendation they expect, and their preferred preference elicitation strategy. Several practical implications are concluded at the end, which might inspire the design and development of effective conversational agents in the news domain. © 2022 Association for Computing Machinery.