Conversational strategies: impact on search performance in a goal-oriented task

Conversational search relies on an interactive, natural language exchange between a user, who has an information need, and a search system, which elicits and reveals information. Prior research posits that due to the non-persistent nature of speech, conversational agents (CAs) should support users in their search task by: (1) actively suggesting query reformulations, and (2) providing summaries of the available options. Currently, however, the majority of CAs are passive (i.e. lack interaction initiative) and respond by providing lists of results – consequently putting more cognitive strain on users.
To investigate the potential benefit of active search support and summarising search results, we performed a lab-based user study, where twenty-four participants undertook four goal-oriented search tasks (booking a flight). A 2x2 within subjects design was used where the CAs strategies varied with respect to elicitation (Passive vs Active) and revealment (Listing vs. Summarising). Results show that when the CA’s elicitation was Active, participant’s task performance improved significantly. Confirming speculations that Active elicitation can lead to improved outcomes for end-users. A similar trend, though to the lesser extent, was observed for revealment – where Summarising results led to better performance than Listing them. These findings are the beginning of, but also highlight the need for, research into design and evaluation of conversational
strategies that active or pro-active CAs should employ to support better search performance.