Sam McBride

@sjamcbride.bsky.social

4348 Followers 24 Following

Northern Ireland Editor of the Belfast Telegraph & Sunday Independent. Author of Burned: The Inside Story of the Cash-for-Ash Scandal. Any views mine alone.

Grok Conversation See new posts Objectively does the X algorithm favour violence and hatred. Be honest. No, the X algorithm does not specifically target or "favour" violence and hatred in the narrow sense of preferring content that explicitly depicts or calls for physical violence over other types. However, it does strongly favour content that provokes intense negative emotions — especially anger, outrage, hostility, and animosity — and this creates a structural bias toward hatred, division, and content that can easily escalate to (or normalise) violent rhetoric. Here is the objective picture based on available evidence: Engagement-first design amplifies outrage — Social media algorithms (including X's) prioritise "conversation-driving" content because likes, replies, quotes, and retweets signal value. Multiple independent studies show anger and out-group animosity are among the strongest predictors of shares/retweets (often 4–6× stronger than other emotions). Content that attacks opponents, mocks groups, or triggers moral outrage reliably gets boosted. X-specific changes since 2022–2023 — After the ownership change, X published parts of its recommendation algorithm. Analyses (including by Amnesty International in 2025) found it systematically prioritises posts that spark heated exchanges, even when false/misleading/harmful. Premium (paid) accounts receive artificial amplification, which helped far-right/racist accounts dominate during high-tension events. Real-world examples of amplification — During the 2024 UK Southport riots, X's "For You" feed rapidly spread false anti-migrant/anti-Muslim claims (despite later police clarification). This contributed to offline violence. Amnesty described it as creating "staggering amplification of hate" because the algorithm rewards provocation over accuracy or safety. Similar patterns appear in other crises (e.g., increased hate speech volume ~50% higher post-2022, with more likes/engagement on hateful posts).
Ken Reid and Stephen Grimason pictured together at Queen's University.
A screenshot of my computer screen...with far too many internet browser tabs open at once.