Bluesky's Block List Problem Is Also a Bias Problem Nobody Wants to Name

A Bluesky user posted something this week that got more traction than its like count suggests: a pointed observation that public block lists — increasingly automated and AI-assisted — might be functioning as engagement hacks rather than safety infrastructure, and that the side effect is discrimination and echo chambers baked directly into how platforms grow. The post didn't go viral. It didn't need to. It landed in a conversation that had been building around a quieter, more uncomfortable version of the AI bias question: not whether algorithms discriminate, but whether the tools built to stop discrimination are themselves doing the discriminating.

The social platform moderation conversation has spent years treating block lists as neutral — user-generated, community-maintained, a democratic antidote to harassment. But as those lists get scraped, aggregated, and increasingly fed into automated systems that pre-filter who sees what, they carry their original biases forward at scale. The Bluesky post names this directly: using public block lists as an engagement hack has negative consequences to user growth and only reinforces discrimination and echo chambers. What makes this observation pointed is that it implicates everyone — the platforms, the safety advocates, and the users who built the lists in good faith. The same logic applies to hiring algorithms that learn from historically biased rejection data, or content moderation models trained on flagged posts from communities that were themselves already over-policed. Bias laundering, dressed up as community safety.

Elsewhere in the conversation, a Bluesky post about AI-mediated hiring made the stakes concrete: with job markets as oversubscribed as they currently are, and AI doing the initial sifting, discrimination against disabled applicants is, as the post put it,