Social network criticised over feature that targets users who are likely to switch to an advertiser’s rival’s product

Facebook has been criticised for hiding an advertising product that claims to be able to predict users’ future behaviours and target messages at them in an attempt to alter those behaviours.

The product, named “loyalty prediction” by Facebook, is part of a suite of capabilities enabled by a machine learning-powered tool called FBLearner Flow. That tool was publicly introduced in 2016, but the advertising techniques it enables were only revealed in a pitch document leaked to the Intercept.

The document, according to the Intercept, describes Facebook’s ability to use its vast amount of personal data to identify individual users who are “at risk” of changing to competitors’ products, and then target them with advertising at the moment they would be about to switch.

“These individuals could then be targeted aggressively with advertising that could pre-empt and change their decision entirely,” writes the Intercept’s Sam Biddle, “something Facebook calls ‘improved marketing efficiency’”.

Tim Hwang, the director of the Harvard-MIT Ethics and Governance of AI initiative, told the publication that Facebook had an ethical obligation to disclose how the technology works.

However, Hwang said: “We can’t forget the history of all this, which is that advertising as an industry, going back decades, has been about the question of behaviour prediction … of individuals and groups … This is in some ways the point of advertising.”

The technology serves to underscore the power of Facebook to profile its users and make predictions about aspects of their offline life.

That capability has led to conspiracy theories about the company listening in to users through their smartphones to better target them for adverts. Although Facebook has repeatedly denied such theories, its chief executive, Mark Zuckerberg, was nonetheless explicitly asked whether it carried out such a practice during his testimony in front of the Senate and House of Representatives last week.

“I’ve heard constituents fear that Facebook is mining audio, which I think speaks to the lack of trust,” Sen Gary Peters said last week. “Does Facebook use audio obtained from mobile devices?” Zuckerberg replied with a flat no – but that did not stop Larry Bucshon, a Republican congressman from Indiana, from bringing up the topic again the following day.

Buschon said he had had a conversation with his mother about a deceased relative, and later that night, Facebook had showed her a memorial video of the relative.

He asked Zuckerberg whether Facebook was contracting with someone else to listen in, and when Zuckerberg replied that the company was not, Buschon pushed, arguing it was pretty clear that someone was listening in.

The questions came despite an explicit denial in 2016 from the social network, in a post titled “Facebook does not use your phone’s microphone for ads or News Feed stories”.

A Facebook spokesperson said: “Facebook, just like many other ad platforms, uses machine learning to show the right ad to the right person. We don’t claim to know what people think or feel, nor do we share an individual’s personal information with advertisers.”