Interpreting algorithmic information cues: User sensemaking of search autocomplete moderation

Abstract Autocomplete is a search feature that algorithmically generates information cues for any keywords entered in the search bar. While this feature makes the search process more efficient, it also frequently produces biased, misleading, offensive, or otherwise inappropriate suggestions. To address this problem, commercial search systems like Google Search now moderate Autocomplete information cues. However, we know relatively little about how users perceive this moderation process. Conducting interviews with 20 users of web search systems, I examine user attitudes toward the ethical tradeoffs in enacting moderation, reliance on search systems for making regulation decisions, and users' own role in moderating autocomplete information. My findings show that users desire greater visibility into the autocomplete moderation process. They also see flags and personal moderation mechanisms as promising avenues for themselves to exert greater agency within contemporary information infrastructures. My analysis bridges the fields of content moderation and search engine critiques, and lays the groundwork for enacting fair, accountable, and transparent search moderation.