automation bias, our tendency to over-rely on automated systems, such as when airline pilots followed incorrect advice from an automated failure-detection system.

This was a project that Google launched in 2008 to predict flu outbreaks by analyzing the search queries that its millions of users make every day. An increase in searches for flu-related terms could be indicative of an imminent outbreak. Google heavily promoted it as an example of AI and mass data collection used for social good. But within a few years, the accuracy of the predictions dropped precipitously. One reason was that it is hard to distinguish between media-driven panic searches and actual increases in flu activity. Another was that Google’s own changes to its app changed people’s search patterns in ways that weren’t accounted for by the AI. Google Flu Trends ultimately ended up as a cautionary tale.