Fish and Sheep from Aofei Temple
Quantum bits | Official Account QbitAI
When 20 million people were engaged in workplace social interaction on LinkedIn , they didn’t know that the platform was quietly using them for experiments.
Five years later, the latest conclusion appears:
Those friends who are not as familiar with you can help you find a job better than your close friends.
research comes from research institutions such as LinkedIn and Harvard Business School , and has been published in Science.
Specifically, they used A/B testing to push different versions of algorithms to 20 million users in the recommended function "Guess You Know".
results found that the most helpful thing for finding a job is often not those friends who interact with you most frequently and are closest in touch with you, but friends with "medium relationships".
pair is the list of people who have about 10 common contacts with you, but they don’t interact very well. Let’s take a look at what’s going on with
.
5-year social experiment for 20 million people
To explain this matter clearly, we have to start with the "weak tie" theory.
973, American sociologist Mark Granovit published a paper titled "The Power of Weak Relationships".This paper is considered one of the most influential sociological papers, with citations reaching nearly 67,000 times.
Granovit proposed in the paper:
The relationship between people and their frequent contact with relatives and friends belongs to a kind of " strong relationship ", and the information obtained through this relationship is often homogeneous.
What is more extensive in society is a kind of interpersonal relationship that is not in-depth. This weak relationship can enable individuals to obtain information that cannot be obtained through strong relationships, thus playing a decisive role in the diffusion of information in work and career.
Simply put, people who are not that familiar with each other often provide information from another social network.
specifically looks at job hunting, which means that job seekers can obtain richer job information from "weak relationships".
This super-large experiment of LinkedIn provides supportive evidence for this theory.
paper author and LinkedIn data scientist Karthik Rajkumar mentioned that in the study, they set up 7 variants of the "Guess You Know" algorithm.
For example, one of the variants will create more connections between users, including more weak relationships. Another variant only recommends people with strong relationships when recommending more contacts to users.
After a large-scale test of 5 years of and
for 20 million users, LinkedIn obtained a large amount of random experimental data. Data analysis results show thaton LinkedIn, 's relatively weak social relationships in promoting employment can achieve twice the stronger social relationships .
△The left picture is the least squares regression analysis
Here "weaker social relations" refers to the situation of having 10 common friends, while "stronger social relations" refers to the situation of having more than 20 common friends.
It should be noted that although in experiments with the number of common friends (above) and the interaction intensity (above) as the relationship evaluation criteria, the least squares regression analysis showed that strong relationships were related to the increase in the probability of changing jobs, the experimental analysis results for a large number of samples, lasting for 5 years, and covering all regions of the world showed the opposite situation.
Finally, the researchers came to three main conclusions: the relatively weak relationship of
- can more increase the possibility of work flow.
- uses the number of common friends as the evaluation criteria. There is an inverted U-shaped relationship between the strength of the relationship and the work flow, that is, the possibility of a medium relationship increasing the work flow is the greatest, and the possibility of a strong relationship increasing the work flow is the least; the evaluation criteria is the intensity of the interaction, and the impact of a strong relationship on the work flow is the greatest.
- Adding relationship nodes with moderate number of common friends and weak interaction is the most beneficial for finding a job.
In addition, the paper also mentioned that the power of weak relationships varies from industry to industry: in industries with higher digitization, weak relationships have stronger power; in industries with lower digitization, strong relationships are more beneficial to finding jobs.
experiment caused controversy
While this experimental result of LinkedIn attracted attention, controversy also followed.
New York Times quoted experts' opinions and pointed out sharply:
LinkedIn's approach may have changed the lives of many people. Performing long-term, large-scale experiments on people may affect their work prospects.
and LinkedIn users are not aware of this experiment.
But some netizens also pointed out that A/B testing is a common means for technology companies to apply algorithms, and conspiracy theories are not necessary. But the key is how big the differences are between the groups.
If the difference is so big as day and night, then LinkedIn does this really immoral. The official spokesperson of LinkedIn also responded to the matter:
We are just trying to make better recommendations for users.
During the test, some people got better algorithms a week or two before others, which would make enough changes in the data so that we could observe and analyze. The object of the experiment is algorithms rather than people.
So, what do you think?
Reference link:
[1]https://www.science.org/doi/10.1126/science.abl4476
[2]https://arstechnica.com/tech-policy/2022/09/experts-debate-the-ethics-of-linkedins-algorithm-experiments-on-20m-users/— End —
Quantum bit QbitAI · Toutiao Sign
Follow us and learn about cutting-edge technology dynamics
first
