Strange it may seem ive never been to pair

We may earn a commission if you buy something from any affiliate links on our site. Learn more.

Facebook's People You May Know algorithm is shrouded in mystery – even within the company itself. But its suggestions have often led to dark consequences

WIRED

For Eddie Velasco, it was a customer at the bank where he worked. For Julice Everage, it was someone in Starbucks. And for Brigid Kaelin, caregiving for her dying mother last summer, it was a nurse who visited her home. Velasco, Everage, and Kaelin are just three of thousands of people who have been creeped out after Facebook’s “People You May Know” tool recommended they connect with strangers shortly after first meeting them in real life.

For over a decade, People You May Know (PYMK) has been recommending potential new friends on Facebook: there’s that girl you knew in primary school, your best friend’s dad, and the uncle you’ve been avoiding adding because of his love of PROUD BREXITEERS AGAINST POLITICAL CORRECTNESS memes. The algorithm behind PYMK is so secretive that it breeds frenzied speculation. If you search for “People You May Know”, the first snippet pulled by Google claims that “People You May Know = People Who Stalked You”, something Facebook itself denies.

Facebook publicly admits there are four factors that feed into PYMK. First there’s people you have lots of mutual friends with – the most common reason for suggestions, according to Facebook’s Help Centre. Then there are people who are in the same Facebook groups or photos as you; people who went to the same school or work at the same company as you; and phone and email contacts you’ve uploaded to Facebook (knowingly or not). While back in 2016, Facebook also said “many other factors” fed into its algorithm, this line has now been deleted from the official Help Centre page.

“My only theory is a ‘tinfoil on the head, they’re watching us’ thing,” says Velasco, a 23-year-old bank teller from Kentucky who recently served a female customer before seeing her on PYMK later that night. Despite multiple investigations proving the contrary, many people still believe Facebook listens in on conversations via your phone’s microphone in order to recommend products or even, in this case, friends.

“I think it connects me to people because Facebook is always listening,” says Everage, a 19-year-old Starbucks barista from Oklahoma who is frequently recommended her customers. Kaelin, a 40-year-old musician from Kentucky, believes her mother’s nurses came up on her suggested friends because their mobile phones were in the same location.

The fog of mystery surrounding PYMK is so thick that even Facebook itself isn’t particularly clear on the issue. “The one thing I have discovered in all my reporting on People You May Know is even the people who work there have little idea how it works, and as a result of that I’ve got very conflicting messages about it over the years,” says Kashmir Hill, a journalist who has reported on PYMK for nearly five years.

In 2016, Hill reported that location services were “one of the factors” Facebook used for PYMK after a company spokesperson confirmed this with her. After her article was published, Facebook retracted the statement – claiming that the company only once ran a four-week test using location services for PYMK.

It’s understandable, then, why people are suspicious, and why rumours persist. In 2016, when Vox ran an article about PYMK, Facebook said it did not collect text and call data from users. Two years later in March 2018, the company admitted that it does collect this data from some Android users via the Messenger app, claiming it did not mislead Vox because the journalists only asked about Facebook, not Facebook Messenger.

Read more: How to delete Facebook

While Facebook now readily admits it uses contact data that you’ve uploaded to feed into PYMK (you can check if you’ve uploaded your contacts to Facebook here or Messenger here), the lack of awareness around this has not only led to conspiratorial thinking, it has also endangered people’s lives and livelihoods.

In 2017, sex workers feared for their safety when PYMK recommended their clients add them on the app. A year earlier, a psychiatrist’s patients were recommended to one another as friends – Hill, who broke this story, theorised that because all of the patients had the psychiatrist’s number saved, Facebook assumed they were a network of connected people. PYMK has, over the years, outed cheating husbands to their wives and recommended rape victims “friend” their rapists. Hill has her own shocking experiences with the feature – in 2017, Facebook connected her with a great aunt who she had never met.

On a daily basis, PYMK causes problems for people, making it remarkable that the feature has no “opt out”. A 22-year-old Redditor from Massachusetts tells me that when Facebook started recommending he add Furries – people interested in andromorphic animals who dress in “fursuits”, sometimes for sexual purposes – he realised one of his friends had the unusual interest. “I was worried sick because my dad was sharing an account with me,” the user says, concerned about sexual images in close proximity to his page. “My friend actually outed himself when I came to him about the Furry invasion.” Thankfully, the friend wasn’t embarrassed or upset.

There are over 300 people on my own People You May Know – when I delete them all, Facebook then recommends 300 more. While Facebook claims “mutual friends” are the most common reason for recommendations, I’m intrigued about how this works in practice. Someone I have zero mutual friends with is my first recommendation, while someone with 16 mutual friends doesn’t appear until halfway down the list. Even though I’ve interacted with this person multiple times, and even been to the pub with them, I don’t particularly like them.

Does Facebook somehow know this? Conversely, a person with only two mutual friends who I’ve never met in real life is right near the top – but I like them a lot, and we’ve been chatting on Twitter. Does Facebook somehow know this too? Have I fed the website enough information over the years for it to understand who I might or might not get along with?

My most common recommendations are people I, as a journalist, have interviewed for stories. What’s unusual about this is that I call interviewees over WhatsApp, and don’t add them as contacts in my actual phone. “WhatsApp does not share contacts with Facebook; we don’t use WhatsApp data to recommend candidates in ‘People You May Know’,” says a Facebook spokesperson when asked about this. Yet, as Hill has pointed out in the past, Facebook could use this data – WhatsApp’s privacy policy says, “Facebook… may use information from us to improve your experiences within their services such as making product suggestions (for example, of friends or connections, or of interesting content)”.

What's most likely is that while I haven’t added my interviewees to my phone address book, they’ve added me, leading to the connection. “What’s so hard about People You May Know is that you can’t guard your own privacy. If someone adds you as a contact, then there’s this leakage that’s happening,” Hill says.

It’s also worth noting that while I personally have been careful not to upload any of my contacts via Facebook itself, Messenger is a totally different story. Until writing this article I was unaware that Facebook was “continuously uploading” my phone contacts to Messenger (here’s how to turn the setting off and, if you’re on Android, ensure Messenger doesn’t also sync your call and text logs).

Mysteries remain about PYMK (Hill has created a tool you can use to help her figure out the algorithm, though naturally Facebook claimed it violated its terms of service). There are a number of potential reasons why Velasco, Everage, and Kaelin saw the recommendations they did. Perhaps their recommended friends were members of the same groups, or had friends from the same school, or were once tagged in pictures together.

Perhaps it’s just a coincidence or an example of the Baader–Meinhof effect – when something (or in this case, someone) who has recently come to your attention appears again with alarming frequency. Perhaps Velasco’s customer was in his PYMK before she came into the bank, but he only noticed her there after?

Still, even if PYMK isn’t creepy enough to check up on your location or listen in on your microphone, the feature still troubles people daily. “I think what’s so creepy about People You May Know is that Facebook could very, very accurately predict who you know in real life and people you’re seeing in real life, because they do have access to all that data, they do have location data on your phone,” Hill says. “They’re just choosing not to use it, as far as we know.”

More great stories from WIRED

– Why parents should stop worrying about videogame addiction

– Why are we having less sex? Blame honesty

– Jeff Bezos wants to colonise space, but he's destroying Earth

– Why staff on the autism spectrum are a huge asset

– Why Tim Cook is a better Apple CEO than Steve Jobs

Amelia Tait is a freelance journalist who writes on culture, trends, and the internet for publications such as The Guardian, The New York Times, Vice, and GQ. She is based in London.