What are the current changes to LinkedIn’s algorithm?
Image Credits:ST.art / Getty Images
LinkedIn’s Gender Bias Experiment: The #WearthePants Initiative
In November, a product strategist, whom we’ll refer to as Michelle (not her real name), logged into LinkedIn and modified her gender to male, adopting the name Michael. This action was part of the #WearthePants experiment, aimed at investigating whether LinkedIn’s algorithm exhibited bias against women.
Engagement Drops Among LinkedIn Users
For several months, frequent LinkedIn users expressed concerns about declining engagement and impressions on the platform. These issues arose after Tim Jurka, LinkedIn’s Vice President of Engineering, revealed in August that the platform had recently started using large language models (LLMs) to enhance content relevance for users.
Michelle, who has over 10,000 followers, noticed a disparity in visibility between her posts and those of her husband, who has only about 2,000 followers. Despite her larger follower count, both achieved similar levels of engagement. “The only significant variable was gender,” she observed.
Significant Changes from Gender Swaps
Another participant in the #WearthePants experiment, entrepreneur Marilynn Joyner, also changed her profile from female to male. She reported a staggering 238% increase in post impressions within just one day. Others, including Megan Cornish and various other women, reported similar results after modifying their profile genders.
LinkedIn’s Response to Bias Claims
LinkedIn has stated that its algorithm does not use demographic factors such as age, race, or gender to determine content visibility. The company contends that differences in engagement do not automatically signify unfair treatment. Social algorithm experts suggest that explicit sexism may not be at play, but implicit biases could be influencing visibility.
Brandeis Marshall, a data ethics consultant, explained that algorithms operate on complex mathematical and social mechanisms. Changing a user’s profile details is just one factor that influences content distribution. She noted, “This is a more complicated problem than people assume.”
The #WearthePants Experiment’s Origin
The #WearthePants initiative was started by entrepreneurs Cindy Gallop and Jane Evans. They had two males create and post the same content as them, seeking to determine whether gender was the reason for perceived drops in engagement among female users. Gallop found her post reached merely 801 people, whereas the male counterpart’s post garnered over 10,000 views, exceeding his follower count significantly.
Joyner, who uses LinkedIn as a marketing platform, expressed concern about the need for LinkedIn to acknowledge any bias within its algorithm.
The Role of Implicit Bias
LinkedIn’s algorithms, much like those of other platforms, have been criticized for embedding a white, male, Western-centric perspective due to the biases of those who train these models. Researchers have identified inherent biases like sexism and racism within these systems.
Despite LinkedIn’s assertions, Marshall highlighted that various unknown factors might account for the rise in impressions after changing gender profiles. Participating in trending topics or users posting after long gaps could both affect visibility.
Tone and Writing Styles Matter
Michelle noted that after transitioning her profile to “Michael,” she adopted a more straightforward, direct writing style, borrowing from her husband’s approach. This audiovisual shift led to a remarkable 200% jump in impressions and a 27% increase in engagement. She concluded that while the system did not appear explicitly sexist, it seemed to value communication styles typically associated with men over those associated with women.
Implications for User Profiles
Sarah Dean, a Cornell assistant professor of computer science, mentioned that platforms like LinkedIn analyze comprehensive user profiles and behavior patterns to determine which content gets amplified. Factors such as job titles and engagement patterns play significant roles in shaping post visibility.
LinkedIn stated their algorithm considers numerous signals from users, including profile data and activity logs. As user behavior evolves daily, engagement metrics continuously reshape what appears in each user’s feed.
Mixed Experiences with LinkedIn’s Algorithm
Many users, regardless of gender, report dissatisfaction or confusion regarding LinkedIn’s updated algorithm. Shailvi Wakhulu, a data scientist, has consistently posted for five years but observed a significant drop in impressions, now achieving only a few hundred views per post.
Another user reported a 50% decline in engagement, even as others claimed a notable increase dependent on topic specificity. This inconsistency indicates the complex nature of algorithmic interactions.
Marshall observed that her content related to her race tends to perform better than posts about her professional expertise. This indicates a potential bias, wherein demographic experiences dictate engagement levels.
Dean opined that the algorithm might simply amplify pre-existing patterns. It may reward content that has historically garnered attention rather than directly correlating engagement rates to the author’s demographics.
Enhancing Transparency and Algorithmic Fairness
LinkedIn acknowledged the growing user base has led to heightened competition, with posting frequency and comments experiencing sizeable year-over-year growth. Specific content types, such as professional insights and industry analyses, are reportedly performing well on the platform.
Despite these observations, users like Michelle demand greater transparency from LinkedIn regarding its algorithm. However, companies often safeguard these secrets as a means to prevent exploitation or manipulation of their platforms, making granular transparency unlikely.
Conclusion
The #WearthePants experiment highlights significant concerns surrounding gender bias and algorithmic fairness within LinkedIn’s content visibility systems. While users strive for equitable treatment and clarity on engagement metrics, the complex nature of these algorithms continues to pose challenges. Solutions will require ongoing research and adjustments to foster a more equitable digital networking experience.
Thanks for reading. Please let us know your thoughts and ideas in the comment section down below.
Source link
#whats #LinkedIns #algo
About The Author
