Do You Know How PYMK or Other Friend-Recommendation Systems Work? Tell Us

Illustration: Jim Cooke/GMG

“People You May Know.” “Suggested Friends.” “Suggestions for you.” No matter what it’s called, most of the social networks have it: a feature that recommends other users for you to connect with.

How do the networks find those other people? Sometimes the recommendations seem creepy and privacy-invasive, like when Facebook suggested a psychiatrist’s patients friend each other, or when it found a long-lost relative, or when it recommended two dads friend each other after they’d attended an anonymous meeting for parents with suicidal teens.

Advertisement

But the suggestions never come with any explanation. They’re produced by hidden algorithms, using invisible data flows. For all the information we give networks about ourselves, they offer almost nothing in return. Facebook says its People You May Know feature, or PYMK, relies on obvious things—mutual friends, shared schools and workplaces, contacts users have shared with them—and on other, unspecified information.

We want to know more about how these networks are keeping track of us. And we hope people who’ve worked on these friend-recommendation features will offer us some insight into how they work.

Have you worked on this technology over the past decade? Do you know how it is able to make such uncanny recommendations? We want to talk to you to learn more.

There are a number of ways to reach us:

Phone/Signal/WhatsApp: 917-999-6143

Email: kashmir.hill@gizmodomedia.com or tips@gizmodomedia.com

Mail:

Gizmodo Media Group
Special Projects Desk
2 West 17th Street
Floor 2
New York, N.Y. 10011

Advertisement

Or send us information securely and anonymously via our Secure Drop: https://specialprojectsdesk.com/securedrop/

Share This Story

About the author

Kashmir Hill

Kashmir Hill is the deputy editor for the Special Projects Desk, which produces investigative work across all of Gizmodo Media Group's web sites. She writes about privacy and technology.

EmailTwitterPosts
PGP Fingerprint: AE77 9CA9 59C8 0469 76D5 CC2D 0B3C BD37 D934 E5E9