Study: LinkedIn's A/B testing of its People You May Know algorithm, involving 4M+ users in 2015 and 16M+ in 2019, could have affected some users' livelihoods
A study that looked back at those tests found that relatively weak social connections were more helpful in finding jobs than stronger social ties. Tweets: @drtechlash , @johncoogan , @alexstamos , @zoeschiffer , @taylorlorenz , @cfchabris , @rmac18 , @natashanyt , @joshelman , @eileenguo , @zenofdesign , @emilygorcenski , @hunterwalk , @mmasnick , @carnage4life , @zck , @jodiecongirl , @kimmytaylor , @as_2103 , @taylorlorenz , @weselymd , @harrybr , @iroughol , @caseynewton , @tomgara , @benjamingoggin , @alexstamos , @pt , @martinsfp , @sub8u , @adamnash , @donmoyn , @hankgreen , @seyitaylor , @eric_seufert , @michaelzimmer , @jeffjarvis , @moonalice , @fabiochiusi , and @natashanyt Tweets: Nirit Weiss-Blatt, PhD / @drtechlash : CONTEXT: Granovetter's “The Strength of Weak Ties” is a seminal social science paper - cited 66,923 times (!) Weak ties have long been proven to contribute to occupational mobility (& higher wages/employment rates). Having LinkedIn prove that on a larger scale is a good thing. https://twitter.com/... John Coogan / @johncoogan : oh damn, i wonder if LinkedIn used the NYT's open-source react tracking package to load their analytics scripts... https://github.com/... https://twitter.com/... Alex Stamos / @alexstamos : I can't wait for the NY Times to discover that when terminally ill patients enroll in experiments that they hope will save them, often half of them get placebos. Making things better and expanding understanding requires hard ethical choices and the Times messed this one up. https://twitter.com/... @zoeschiffer : “But the changes made by LinkedIn are indicative of how such tweaks to widely used algorithms can become social engineering experiments with potentially life-altering consequences for many people” https://www.nytimes.com/... Taylor Lorenz / @taylorlorenz : These tests run by LinkedIn are indicative of how such tweaks to widely used algorithms can become social engineering experiments with potentially life-altering consequences for many people. https://www.nytimes.com/... Christopher Chabris / @cfchabris : In other news, LinkedIn ran thousands of experiments by constantly changing features of their site & product without scientifically testing them, which affected people's lives and made money for the company, but we know nothing about any of their effects https://www.nytimes.com/... Ryan Mac / @rmac18 : This shady experiment run on LinkedIn using the “People You May Know” algorithm reminds me of a story I tried to report involving famous accounts that I could never prove out, but always wondered about... https://www.nytimes.com/... Natasha Singer / @natashanyt : A fascinating new study in Science details how LinkedIn ran social experiments on 20 million users over 5 years. It shows how algorithmic tweaks can become social engineering experiments with potentially life-altering consequences for millions of people. https://www.nytimes.com/... Josh Elman / @joshelman : There are so many ways to interpret this story. The most optimistic one is that LinkedIn has helped many people get new jobs and advance their careers and lives. And they continue to try to make it better and learned an important new way to help people https://www.nytimes.com/... Eileen Guo / @eileenguo : The thing is that companies use users to test and train their AI all the time — and we only know about LinkedIn's latest because it resulted in a peer reviewed study. When it's just for internal “product improvement,” we have no idea. https://www.nytimes.com/... @zenofdesign : Boy, if you hate this, you'll hate every major website and mobile game made in the last ten years. https://twitter.com/... Emily G / @emilygorcenski : Second verse same as the first https://zpr.io/... @hunterwalk : when the goal of the test is POSITIVE [can we help our users find better outcomes] the framing of routine A/B testing as nefarious in its lack of disclosure is quite cynical https://twitter.com/... Mike Masnick / @mmasnick : I am confused why standard every day a/b testing that basically every internet company always does is a NY Times story? https://twitter.com/... Dare Obasanjo / @carnage4life : An A/B test to hide minority candidates from recruiters is unethical even if some recruiters prefer it. On the other hand, a test to see if alumni from your college or your former employer are more likely to connect with you isn't. It's simply how you improve the feature. Zak Kukoff / @zck : Shot, chaser https://twitter.com/... https://twitter.com/... Jodi Beggs / @jodiecongirl : meanwhile i get questions whenever i want to run an experiment for more than 4 weeks so rather than being mad i'm pretty jealous tbh https://twitter.com/... Kim Taylor / @kimmytaylor : Today the NYT discovers A/B testing https://twitter.com/... @as_2103 : @mahaniok ... Unethical as per the business terms. Many companies ask the customers to sign up for beta programs to keep these kind of ethical issues at bay. It shouldn't be a big deal for a platform like LinkedIn to get 20M subscribers for beta program. Taylor Lorenz / @taylorlorenz : “The findings suggest that some users had better access to job opportunities or a meaningful difference in access to job opportunities. These are the kind of long-term consequences that need to be contemplated...” https://www.nytimes.com/... @weselymd : These data are interesting. Contacts w <10 mutual “friends” proved more helpful to people in finding jobs. Makes sense as that's how we “expand our territory” in terms of connections to new circles. LinkedIn is not my jam but this reminds me to get out of my comfort zone. 🤙 https://twitter.com/... Harry Brignull / @harrybr : Speaking an ex academic - all Ψ experiments go via an ethics committee, and you have to comply with loads of rules, that include making sure that participants understand. It's ok to be upset about companies running AB tests on you without your consent. https://www.nytimes.com/... Isabelle Roughol / @iroughol : Absurd writing a “look what we found out” headline when the facts come out of a press release from the company and open social science research that was widely publicised. https://twitter.com/... Casey Newton / @caseynewton : LinkedIn also ran an experiment on me. It sent me 48,000 emails designed to get me to visit LinkedIn https://www.nytimes.com/... Tom Gara / @tomgara : The NYT dialed up the sensationalism on this, but also, if if I didn't want my A/B testing to be described as a massive social experiment, I would simply not publish the results as a breakthrough piece of social science research in a prestigious science journal https://twitter.com/... Benjamin Goggin / @benjamingoggin : LinkedIn experimented with millions of users, improving job prospects for some https://www.nytimes.com/... Alex Stamos / @alexstamos : Universities have detailed, rigorous processes in place to weigh the difficult ethics of improving the world via human subject research. This paper was co-authored by MIT, Stanford and Harvard faculty. The MIT IRB approved the protocol, which the Times completely omits. https://twitter.com/... Parker / @pt : I genuinely wonder if the author of this article on the ethics of AB testing has any idea that the NYT even does it, given they don't disclose it anywhere. I'm sure AB testing the paper of record for engagement is less of a risk to society than friend suggestions tho, so NBD? https://twitter.com/... @martinsfp : It's easy to stay this article is over-egging the impact a bit, but it certainly highlights that second-order impacts of A/B tests on social platforms should be considered more. https://www.nytimes.com/... @sub8u : Looking forward to receiving a gazimillion notifications from NYT and every other web service globally for when they A/B test every other headline, article placement, story recommendation & take explicit approval. Better still, in facsimile, in triplicate. https://www.nytimes.com/... https://twitter.com/... Adam Nash / @adamnash : Running experiments is how we learn things. Hard to imagine how we'd know whether our ideas to improve product were effective without it. Scientific method. Much better than rolling out changes without testing! This piece from the @nytimes is 🤯 https://www.nytimes.com/... Don Moynihan / @donmoyn : Part of the story here is that it is bad for social media companies to do research on how users experience the platform. But it's better to have that research than not. https://www.nytimes.com/... Hank Green / @hankgreen : Video game companies experiment on children with slot-machine mechanics to determine what ratio of loot box winning results in kids gambling more and we ignore it, but LinkedIn A/B testing a feature is...news??? https://twitter.com/... @seyitaylor : this is quite literally their job. https://twitter.com/... Eric Seufert / @eric_seufert : I'm a bit alarmed at seeing a routine A/B test regimen characterized as a “social experiment” for which explicit, opt-in consent should have been obtained. https://www.nytimes.com/... Michael Zimmer / @michaelzimmer : LinkedIn ran experiments on more than 20 million users over five years. I comment on the ethical consequences of this type of big data research in this great article by @natashanyt in the @nytimes. https://www.nytimes.com/... Jeff Jarvis / @jeffjarvis : Demonizing A/B testing. Without this “social experiment” (loaded phrase) they would not know how to improve prospects for all. LinkedIn Ran Social Experiments On 20 Million Users Over Five Years https://www.nytimes.com/... Roger McNamee / @moonalice : This story is a reminder that Microsoft is major player in surveillance capitalism (they harvest data from childhood to adulthood: Minecraft>Xbox>LinkedIn> search). Now there is evidence they experimented on 20 mm LinkedIn users, potentially affecting their careers. https://twitter.com/... Fabio Chiusi / @fabiochiusi : “LinkedIn ran experiments on more than 20 million users over five years that, while intended to improve how the platform worked for members, could have affected some people's livelihoods, according to a new study” https://www.nytimes.com/... Natasha Singer / @natashanyt : Remember how Facebook experimented on 700K users for a weeklong study looking at how people's emotions spread on the platform? LinkedIn experimented on 20 million users over 5 years by tweaking their People You May Know suggestions. https://www.nytimes.com/...
Context & Ripple Effects
The study behind this story is a retrospective audit of LinkedIn's own logs: between 2015 and 2019 the company ran People You May Know experiments on roughly 20 million users without their knowledge, and the researchers found that the weak ties those tests surfaced did more for job-seekers than strong connections would have. That means LinkedIn wasn't just tuning a feed — it was quietly shaping who heard about which jobs.
The finding lands in a corpus where platform experimentation has been repeatedly re-examined after the fact: Meta later co-published four studies conceding its algorithm is 'influential' but hard to steer, while Facebook's earlier firing of 52 staffers for data misuse forced an internal governance overhaul. LinkedIn's case adds a sharper edge — the harm claimed isn't belief-shifting or privacy loss but livelihoods.
First-order effects
- LinkedIn now faces questions about why experiments touching 4M+ users in 2015 and 16M+ in 2019 ran without user awareness, putting its research-ethics and consent practices under the same scrutiny previously aimed at Facebook's data handling.
- Job-seekers whose opportunity networks were determined by the tested algorithm are the directly affected population — the study reframes PYMK from a convenience feature into a gatekeeper for employment access.
Second-order effects
- Rival professional networks and recruiters gain an argument that human-curated introductions outperform algorithmic weak-tie suggestions, pressuring LinkedIn to justify the feature's design choices publicly.
- Academic-platform collaborations get riskier: after Meta's co-authored studies drew attention to what platforms can and can't demonstrate about their own algorithms, researchers may push harder for access to experiment logs like LinkedIn's — access companies control unilaterally.
Third-order effects
- If retrospective audits of platform A/B tests keep surfacing real-world stakes, expect calls for independent review-board oversight of large-scale social experiments — moving platform testing from an internal product function toward something resembling regulated human-subjects research.
- The broader pattern across these cases — CubeYou's quiz apps, Cambridge Analytica's models, LinkedIn's PYMK tests — points toward treating social-graph inference as consequential infrastructure, with disclosure obligations attached to who gets recommended to whom.
The trend: Platform-run experiments on users are being retroactively audited as social science with real-world stakes, pushing recommendation systems toward external ethical scrutiny rather than purely internal governance.