Algorithmic bias and fairness

LinkedIn's public position on bias in the Feed is that it works to find and address unfair bias in anything it builds, including the AI models that recommend and rank content1. LinkedIn says it does not limit how widely content is shown, or remove it, based on race, ethnicity, political views or other characteristics1. Different but similar posts can still get different engagement, because the Feed weighs hundreds of other things it looks at, including your position or industry, your network and your activity4.

LinkedIn's work to address bias includes reviewing the ethics and compliance of AI-powered products in its Trustworthy Design Review, a process it applies to product initiatives before launch23. LinkedIn checks its algorithms for bias between two gender categories, using gender that members provide themselves (called Self-ID) together with gender that LinkedIn estimates in two categories23. Members can report job posts they believe may be discriminatory for LinkedIn to review; once a job post is found to be discriminatory, LinkedIn removes lookalike postings, meaning other job posts that closely match it3.