LinkedIn may have hurt job seekers after algorithm testing

THE NEW YORK TIMES – The LinkedIn conducted experiments with more than 20 million users over 5 years which, although aimed to make the platform work better for those who use it, may have affected the working lives of some people, according to a new report. study.

In experiments around the world from 2015 to 2019, Linkedin randomly changed the ratio of weak contacts to strong contacts suggested by the “people you might know” algorithm – it’s the company’s automated system for recommend new connections to its users. Researchers from LinkedIn, Massachusetts Institute of Technology (MIT), Stanford University and Harvard Business School then analyzed the aggregate test data in a study published this month in the journal Science.

LinkedIn’s experiments with algorithms may surprise millions because the company did not notify users that the tests were in progress.

Tech giants like LinkedIn, the world’s largest professional social network, regularly run large-scale experiments in which they test different versions of app functionality, design for web versions, and algorithms with various groups of people. The long-standing practice, called A/B testing, aims to improve the consumer experience and keep them engaged, which helps companies make money from premium subscriptions or advertising. Users often have no idea that they are participating in company-led experiments with them.

But the changes adopted by LinkedIn show how such tweaks to widely used algorithms can become experiments in social engineering with life-altering consequences for many people. Experts who study the societal impacts of computing have said that carrying out large-scale, long-term experiments with people – and which could affect their employment prospects in ways that are not obvious to them – raised questions about industry transparency and research oversight. . .

These are the kinds of long-term consequences that need to be considered when thinking about the ethics of participating in this type of big data research.

Michael Zimmer, professor at Marquette University

Continue after ad

“The results suggest that some users had better access to job postings or a significant difference in their access,” said Michael Zimmer, professor of computer science and director of the Data, Ethics and Society Center at Marquette University. “These are the kinds of long-term consequences that need to be considered when thinking about the ethics of participating in this kind of big data research.”

The study published in Science tested a famous theory in sociology called “The Strength of Weak Ties”who argues that people are more likely to get jobs and other opportunities through distant acquaintances than close friends.

Researchers examined how changes to LinkedIn’s algorithm affected users’ career mobility. They found that relatively weak social connections on LinkedIn were twice as helpful in getting a job as stronger social connections.

In a statement, LinkedIn said that during the study it “acted consistently” with the company’s terms of service, privacy policy and user settings. The privacy policy mentions that LinkedIn uses users’ personal data for research purposes. The release also says the company used the latest “non-invasive” social science techniques to answer important research questions “without any user experimentation.”

Microsoft bought LinkedIn for $26.2 billion in 2016
Microsoft bought LinkedIn for $26.2 billion in 2016

LinkedIn, which is owned by Microsoft, did not provide a direct answer when asked how the company assessed the potential long-term consequences of its experiments on users’ jobs and economic status. But the company said the survey did not disproportionately favor some of them.

The goal of the research was “to help people on a large scale,” said Karthik Rajkumar, an applied research fellow at LinkedIn and one of the study’s co-authors. “No one was harmed while looking for a job.”

Continue after ad

Sinan Aral, professor of management and data science at MIT and lead author of the study, said the LinkedIn experiments were an attempt to ensure users had equal access to job postings.

“Running an experiment with 20 million people and then releasing a better algorithm for everyone’s job prospects based on what they learned is what they’re trying to do,” Aral said, ” it is not a question of choosing a few people to have social mobility or not”.

For experiments, LinkedIn adjusted its algorithm to randomly vary the prevalence of strong and weak links recommended by the system. The first wave of testing, conducted in 2015, “had more than 4 million research subjects”, according to the study. The second wave of tests, carried out in 2019, involved more than 16 million people.

In testing, those who clicked on the “People You May Know” tool and analyzed the recommendations received different algorithmic paths. Some of these “processing variants,” as the study calls them, caused LinkedIn users to connect more with people with whom they had only weak social ties. Various tweaks have caused others to create fewer connections with weak links.

It is unclear whether most LinkedIn users understand that they may be search subjects for experiences that may affect their employment opportunities.

LinkedIn’s privacy policy says the company may “use the personal data we have” to research “workplace trends, such as the availability of jobs and the skills needed for those jobs.” The company’s policy for external researchers who wish to analyze data from the platform clearly states that these researchers will not be able to “perform experiments or tests with our users”.

Continue after ad

But none of the policies explicitly tell users that LinkedIn itself might experiment or test them.

In a statement, LinkedIn said, “We are transparent with our users through the research section of our Terms of Service.”

In an editorial statement, the Science said: “Our understanding, and that of the reviewers, was that the experiments conducted by LinkedIn were conducted within the guidelines of their Terms of Service.”

After the first wave of tests with the algorithms, researchers from LinkedIn and MIT came up with the idea of ​​analyzing the results of these experiments to evaluate the weak tie strength theory. While the theory that’s been around for decades has become a mainstay of social science, it hasn’t been rigorously proven in a large-scale prospective study that randomly assigned people to social bonds with different strengths.

External researchers analyzed aggregate data from LinkedIn. According to the study, people who received more referrals from moderately weak contacts tend to apply for more jobs and be hired more – findings that coincide with weak tie theory.

In fact, relatively weak contacts – that is, people with whom LinkedIn members shared only 10 mutual connections – proved to be much more successful in finding jobs than stronger contacts with whom users had more than 20 mutual connections, according to the study.

Continue after ad

One year after logging on LinkedIn, people who received the most referrals for moderately low contacts were twice as likely to get jobs at companies where those acquaintances worked as those who received the most referrals for strong bonds.

“We’ve found that these moderately weak ties are the best option for helping people find new jobs and more than stronger ties,” said LinkedIn researcher Rajkumar.

The 20 million users involved in LinkedIn’s experiments have created more than two billion new social connections and applied for more than 70 million jobs, leading to 600,000 new jobs. Weak ties were found to be more useful for job seekers in digital fields such as artificial intelligence, while strong ties were found to be more beneficial for jobs in industries less reliant on software, the study found. .

LinkedIn said it applied the weak tie findings to a number of resources, including a new tool that notifies users when a first- or second-degree connection is recruiting. But the company made no study-related changes to its “People You May Know” feature.

Aral, a professor at MIT, said the study’s greatest merit was that it showed the importance of powerful social media algorithms – not only for amplifying issues like misinformation, but also as key indicators of economic conditions. such as employment and unemployment.

Catherine Flick, senior researcher in IT and social responsibility at De Montfort University in Leicester, England, described the study as a new exercise in corporate marketing.

“The study has an inherent bias,” he said. “It shows that if you want to access more jobs, you should use LinkedIn more.”

Add Comment