That’s Not Me: Surprising Algorithmic Inferences
by: Hautea, S., Munasinghe, A., and Rader, E.
Online platforms such as Google and Facebook make inferences about users based on data from their online and offline behavior that can be used for various purposes. Though some of these inferences are available for users to view, there exists a gap between what platforms are actually able to infer from collected data and what inferences users are expecting or believe to be possible. Studying users’ reactions to inferences made about them, espe- cially what surprises them, allows us to better understand this gap. We interviewed users of Google and Facebook to learn their current beliefs and expectations about how these platforms use their data to make inferences, and identified four common sources of surprise for participants: irrelevant inferences, outdated inferences, inferences with no connection to online activity, and inferences related to friends or family. We discuss the implications for designing inference-generating systems.
Hautea, S., Munasinghe, A., and Rader, E.. “That’s Not Me: Surprising Algorithmic Inferences.” Poster in Extended Abstracts of the 2020 CHI Conference On Human Factors In Computing Systems. 2020.