OK, what’s going on with LinkedIn’s algo?

In a novel experiment dubbed #WearthePants, a product strategist—referred to as Michelle for anonymity—decided to switch her LinkedIn profile from female to male. She did this to investigate claims about the platform's algorithm potentially favoring male users. For several months, numerous LinkedIn users had been vocal about a noticeable decline in engagement and impressions on their posts, raising questions about the platform's fairness. This scrutiny followed comments from Tim Jurka, LinkedIn’s Vice President of Engineering, who mentioned in August that the company had recently integrated large language models (LLMs) to enhance user content visibility. Michelle, who boasts over 10,000 followers, noticed that her posts received similar engagement levels to her husband’s, who has only around 2,000 followers. The only significant difference? Gender. Another participant in the experiment, Marilynn Joyner, reported a staggering 238% increase in post impressions after changing her profile gender from female to male. Other women, including Megan Cornish and Jessica Doyle, shared similar experiences, triggering a broader discussion about potential algorithmic bias on LinkedIn. In response to these claims, LinkedIn asserted that its algorithm does not factor in demographic information like age, race, or gender when determining content visibility. The company emphasized that variations in feed engagement do not inherently imply bias. However, experts in social algorithms suggest that while overt sexism may not be at play, implicit biases could still influence outcomes. Brandeis Marshall, a data ethics consultant, explained that the algorithm operates through a complex interplay of various factors, including user interactions. She pointed out that changing a profile name or image is just one element that might impact visibility. The broader implications of these findings led to speculation that certain communication styles stereotypically associated with men might be favored by the algorithm, reflecting a subtle bias. The #WearthePants initiative was introduced by entrepreneurs Cindy Gallop and Jane Evans, who wanted to determine whether gender was a contributing factor to the declines in engagement experienced by women. Their inquiry revealed that their posts garnered significantly fewer impressions than those of male counterparts, despite similar content. LinkedIn maintains that it rigorously tests millions of posts to ensure equitable user experiences. The platform indicated that demographic data is utilized solely for testing purposes to evaluate content visibility among different creators. Nonetheless, as users continue to express frustration with the algorithm, the complexity of its operation presents challenges in identifying the exact causes of engagement fluctuations. Chad Johnson, a LinkedIn sales expert, noted that the importance of traditional metrics like likes and comments has diminished under the new algorithm. Instead, the focus seems to be on the clarity and value of the content being shared. This shift has left many users perplexed about the best strategies for maximizing their reach on the platform. Despite LinkedIn’s assurances, the conversation surrounding algorithmic bias is far from resolved. Users, particularly women, are calling for greater transparency regarding how the platform’s AI systems function. As the debate continues, it remains evident that LinkedIn's algorithm is a topic of significant concern among its diverse user base, impacting how content is created and shared across the platform.