Google Flu Trends

Google Flu Trends ('GFT') was a service that attempted to accelerate the prediction of flu outbreaks in over 25 countries using daily algorithmic analysis of 40 search terms associated with flu such as 'cough' or 'fever' on Google's default search engine.

Released in 2008, GFT was discontinued in 2015 after criticism that it was inaccurate and potentially an abuse of privacy of Google search engine users.

Risks and harms

Accuracy, reliability

Google initially claimed (pdf) GFT was 97% accurate, compared with US Centers for Disease Control and Prevention (CDC) data. But reports and research, notably a Nature article and Harvard study (pdf), concluded the tool over-estimated flu cases and the number of doctor visits likely to occur. 

For example, GFT was found to have overestimated the number of doctors’ visits by over 50 percent in 2011/2012.

Privacy, security

According to Google, GFT could 'never be used to identify individual users because we rely on anonymized, aggregated counts of how often certain search queries occur each week.' 

However, citing research by Carnegie Mellon University computer science professor Dr Latanya Sweeney, showing the 're-identification' of anonymous data is possible, US privacy advocacy organisation EPIC warned (pdf) then Google CEO Eric Schmidt that it could provide a way for government agencies and others to identify Google users.

Transparency 

Google was criticised for providing inadequate data or information about its data sources for external researchers to conduct meaningful analysis. 

Research, advocacy

• Lazer D., Kennedy R., King G., Vespignani A. (2014). The Parable of Google Flu: Traps in Big Data Analysis (pdf)

• Butler D. (2013). When Google got flu wrong

• EPIC (2008). Letter to Google CEO Eric Schmidt (pdf)

• Malin B, Sweeney L., Newton E. Trail Re-identification: Learning Who You are From Where You Have Been