LinkedIn conducted an experiment with more than 20 million users over a five-year period, and a new study found that while it was aimed at improving how the platform works for its members, some It may have affected people’s lives.
In experiments conducted around the world from 2015 to 2019, Linkedin compared weak and strong contacts suggested by its “People You May Know” algorithm, its automated system for recommending new connections to users. Randomly changed the ratio.Tested in detail in research It was published this month in the journal Science and co-authored by researchers from LinkedIn, MIT, Stanford, and Harvard Business School.
LinkedIn’s algorithm experiment may surprise millions because it didn’t notify users that the test was underway.
Tech giants like LinkedIn, the world’s largest network of experts, routinely run large-scale experiments to test different versions of app features, web designs, and algorithms with different people.A/ A long-standing practice called B-testing aims to improve the consumer experience and keep them engaged, helping businesses make money through premium membership fees and advertising. often you don’t realize you’re running (New York Times use a test like that To evaluate headline language and make decisions about the products and features the company releases. )
But the changes LinkedIn made show that such tweaks to a widely used algorithm can be a social engineering experiment that could change many people’s lives. Experts who study the social impact of computing have made industry transparent by conducting long-term, large-scale experiments on people that could affect their job prospects in ways they can’t see. It raised questions about sex and research oversight.
“The findings suggest that some users have better access to employment opportunities, or that there are significant differences in access to employment opportunities,” he said. Michael Zimmer, Associate Professor of Computer Science and Director of the Center for Data, Ethics and Society at Marquette University. “These are the kind of long-term consequences that need to be considered when considering the ethics of engaging in this kind of big data research.”
A study in Science tested an influential theory in sociology. “The strength of weak tiesIt argues that people are more likely to get employment and other opportunities through arm’s length acquaintances than through close friends.
Researchers analyzed how changes to LinkedIn’s algorithm affected users’ job mobility.they Social ties were found to be relatively weak Securing jobs on LinkedIn proved to be twice as effective as strengthening social ties.
In a statement, Linkedin said it “consistently followed” its user agreement, privacy policy and member settings during the investigation.of privacy policy Please note that LinkedIn uses members’ personal data for research purposes. The statement added that the company uses the latest “non-invasive” social science technology to answer important research questions “without conducting experiments on its members.”
Microsoft-owned LinkedIn did not directly answer questions about how the company considered the potential long-term effects of the experiment on its users’ employment and economic status. He said he did not give department users an unfair advantage.
The goal of this research was to “help people at scale.” Karthik RajkumarAn Applied Research Scientist at LinkedIn and was one of the co-authors of this study. “No one was put at a disadvantage in finding a job.”
Sinan Aral, lead author of the study and a professor of management and data sciences at the Massachusetts Institute of Technology, said LinkedIn’s experiment is an effort to give users equal access to employment opportunities. .
“What they’re trying to do is run an experiment with 20 million people and, as a result of the knowledge they’ve learned, deploy better algorithms for everyone’s job prospects.” Professor Alal “Instead of anointing some people to have social mobility and others not,” he said. (Professor Allal did a data analysis for The New York Times, Research Fellowship Grant From Microsoft in 2010. )
User experimentation by major internet companies has a complicated history. 8 years ago, Facebook survey A social network has published a description of how it silently manipulated posts appearing in users’ news feeds to analyze the spread of negative and positive sentiment on its platform.689,003 people The week-long experiment, conducted with users of
A Facebook study that included authors from the company’s researchers and a professor at Cornell University claimed that people tacitly consented to emotion manipulation experiments when they signed up for Facebook. “All users consent before creating an account on Facebook,” the study said, “constituting informed consent for this study.”
Critics disagreed, with some accusing Facebook of invading people’s privacy and exploiting people’s moods to cause emotional distress. Others argued that the project used academic co-authors to lend credibility to the company’s research practices in question.
Cornell later said that Facebook conducted its own research and that professors who helped design the study not directly involved in experiments on humans.
LinkedIn’s professional networking experiments differed in intent, scope, and scale. They were designed by Linkedin as part of the company’s ongoing effort to improve the relevance of its “People You Know” algorithm that suggests new connections to its members.
The algorithm analyzes data such as members’ work histories, job titles, and connections with other users. We then attempt to measure the likelihood that a LinkedIn member will send a friend invite to a proposed new connection and the likelihood that the new connection will accept the invitation.
For the experiment, LinkedIn adjusted the algorithm to randomly vary the prevalence of strong and weak ties that the system recommends. The first wave of tests, conducted in 2015, “had over 4 million subjects,” the study reports. More than 16 million people participated in the second wave of tests in 2019.
During testing, people who clicked on the “People You May Know” tool and saw the recommendations were assigned to different algorithmic paths. Some of these “therapeutic variants,” as the study called it, caused LinkedIn users to form more connections with people with weaker social ties. I started to form fewer connections.
It is unclear whether most LinkedIn members realize that they may be subject to experiments that may affect their employment opportunities.
LinkedIn’s privacy policy The company says it may “use available personal data” to study “workplace trends such as job availability and the skills required for these jobs.”this is Policy for External Researchers Researchers who attempt to analyze the company’s data explicitly state that those researchers cannot “experiment or test our members.”
However, neither policy explicitly informs consumers that LinkedIn itself may run experiments or tests on its members.
In a statement, LinkedIn said, “Through the research section of our User Agreement, we are providing transparency to our members.”
In an editorial statement, Science said:
After the first wave of algorithm tests, researchers at LinkedIn and MIT came up with the idea of analyzing the results of these experiments to test the theory of weak-tie strength. Decades ago the theory was a cornerstone of the social sciences, but it had not been rigorously proven in large prospective trials that randomly assigned people to social ties of varying strengths.
External researchers analyzed aggregated data from LinkedIn. The study reported that people who received more referrals for moderately weak contacts generally applied for and accepted more jobs.
In fact, relatively weak contacts whose LinkedIn members shared their interconnections only 10 times were far more productive in their job search than strong contacts whose users shared their interconnections 20 or more times. It was proved that there is
After a year of connecting on LinkedIn, people who were recommended a moderately weak connection were twice as likely to land a job at the company that the acquaintance worked for, compared to other users who were recommended a strong connection. I was.
Linkedin researcher Rajkumar said:
The 20 million users who participated in the LinkedIn experiment made more than 2 billion new social connections, completed more than 70 million job applications, and led to 600,000 new jobs, the study reported. I’m here. Weak ties proved most useful for job seekers in digital fields such as artificial intelligence, while strong ties proved more useful for employment in industries less dependent on software. Research says.
LinkedIn said it has applied its findings on weak ties to several features, including new tools. Notify members If primary or secondary connectivity is employed. However, the company hasn’t made any research-related changes to the “People You Know” feature.
Professor Allal of the Massachusetts Institute of Technology said the deeper implications of the study were that it showed the importance of powerful social networking algorithms. This not only amplifies problems such as misinformation, but is also important as a fundamental indicator of economic conditions such as employment and unemployment.
Catherine Frick, a senior researcher in computing and social responsibility at De Montfort University in Leicester, England, describes the study as something of a corporate marketing exercise.
“There is an inherent bias in this study,” Dr. Flick said. “This shows that if you want more jobs, you should use LinkedIn more.”