Examining how search engine users understand the production of Autocomplete suggestions

Autocomplete is a popular search feature that automatically generates query suggestions for any keywords entered in the search bar. In this research, I examine regular end-users’ folk theories of how general-purpose search engines produce such suggestions. Drawing on interviews with 20 search engine users, I found that users conceptualize Autocomplete as an automated agent that is influenced by three main factors: (1) searcher’s personal search history and profile, (2) aggregate population-wide queries, and (3) commercial advertising. Users’ assumption of these influences raises for them critical concerns about privacy, transparency, information insularity, targeted data manipulation, and the reproduction of societal biases in Autocomplete’s outputs. My analysis also shows that users view explanations as a promising mechanism to enhance accountability in Autocomplete systems. I highlight the factors that shape users’ mental models of Autocomplete and discuss how their algorithmic imaginaries stabilize platforms’ revenue models.