Voice-Based Proactive Information Retrieval can support social interactions by augmenting
conversations and removing the need for explicit search activity. Previous work introduces the
SearchBot, a proactive search agent used to collect data about search behaviour during social
interactions. Although prior analyses show that it positively influences the conversations of its
users, the research leaves a gap in understanding how it affects their other behaviours. This
thesis aims to bridge this gap by analyzing data from a previous study and characterizing the
influence of the SearchBot on the behaviours and activities of its users. Our findings show that
study participants displayed an increased frequency of engagement with the SearchBot system
than with a more traditional search system. In addition, our exploration of the different types
of search activities that users perform shows that SearchBot users are able to avoid the most
cognitively expensive one (query formulation and typing). The findings also reveal patterns of
interaction between the SearchBot system and its users in terms of speech patterns and search
behaviours. We discuss the implications of our findings and provide suggestions for future work.
