WE HAVE STOPPED DISAGREEING.
WE NOW DISLIKE EACH OTHER.

Algorithmic feeds + the collapse of local news + cable-news outrage cycles produced a media environment where engagement = anger. The platforms didn't invent partisan hostility; they industrialised it. The political effects are downstream of an attention-economy business model that was never aligned with civic life.

“The first amendment defends against censorship. It does not defend against algorithmic amplification.”

Over decades, political identity in the US (and increasingly Europe) has aligned with racial, religious, urban-rural, and educational identities. The result: any political loss now feels like an identity loss, and out-party hostility looks like in-group defence rather than disagreement.

Political elites polarised first; voters followed. Donor classes, primary electorates, and ideological media gatekeepers reward purity and punish compromise. Civility norms eroded at the top before they eroded at the bottom; the fix has to start there too.

Decades of stagnant working-class real incomes + visible elite enrichment create a baseline of resentment. Demagogic politics is a predictable response, not a deviation. Until the material conditions change, asking citizens to be more civil to each other will fail.

Where shared religious life, fraternal orders, and civic clubs have collapsed, political identity rushed in to fill the meaning vacuum. The hyperpoliticisation we see is partly the cost of a society that ran short of binding institutions older than the partisan divide.

Polarisation looks different in India (Hindutva vs minorities), Brazil (post-Lula urban-rural), the Philippines (drug-war moral panic), Hungary (illiberal nationalism), and Turkey (Erdoğan personalism). The pattern is global but the local mechanisms are distinct. Diagnoses imported from US-centric frameworks travel badly.

Structured cross-partisan deliberation reliably reduces out-group hostility in randomised trials. Deliberative Polling, America in One Room, Hidden Common Ground. Effects are real and persistent at small scale; scaling them politically is the unsolved problem.

Many of the harms are downstream of design choices. Algorithm transparency, chronological-feed defaults, friction on virality, bans on rage-engagement micro-targeting. The EU Digital Services Act is the first major attempt to enforce some of this; the early evidence is mixed but the precedent matters.

Counties that lose their local newspaper see measurable spikes in straight-ticket voting and polarisation within a few years. Subsidising / endowing local journalism is one of the highest-leverage civic interventions there is, both upstream of polarisation and downstream of accountability.

Ranked-choice voting, jungle primaries, independent redistricting, multi-member districts. Each shifts the median-voter calculus toward broad coalitions and away from primary-purity extremism. The mechanism design literature is far ahead of the political will.

Most measured ways to lower out-group hostility involve being in regular non-political contact with people unlike you. Churches, mosques, temples, sports leagues, parent-teacher organisations, neighbourhood councils. Re-building these is slow but it is the durable answer.