Paul Worthington wrote an article for Off Kilter titled Cultural Pluralism (25 Sept 2026).
« Culture appears to be fragmenting and experimenting faster than ever at the edges while becoming increasingly stagnant at the center… This is because we democratized production and distribution, but not the commercial mechanisms that migrate culture from the edge toward the center. »
Worthington wonders « whether the more interesting creative opportunity for AI isn’t generating more creative output at all, but redesigning the cultural distribution systems through which creativity is discovered, rewarded, and scaled »
« What if AI, as probabilistic software programmed through natural language, allowed us to build systems designed to seek uncertainty rather than eliminate it; to surface distant adjacencies, anomalies, and things that may sit well outside of what platforms already know we like? »
« Not another personalized feed optimized to exploit our attention ever more precisely, but something almost the opposite: a computational system that feeds our curiosity with genuinely different things worth exploring. »
« Pluralistic elitism. The old media world was hardly a bastion of democratized access. A&Rs, editors, gallery owners, commissioners, studio executives, etc., all had enormous power to decide what the rest of us heard, saw, and experienced. »
« Algorithmic orthodoxy. Today’s cultural gatekeepers … TikTok, Instagram, YouTube, Spotify, LinkedIn… Creators copy what succeeds… Sociologists call such convergence isomorphism… The problem isn’t commercial exploitation itself; it’s that our exploitation systems have become so extraordinarily efficient at extending proven forms they’re crowding out exploration as a viable pathway to success. »
« First, we need to restore cultural pluralism. The old system was bottlenecked by a small number of tastemakers, while the current system is bottlenecked by algorithmic convergence. While dropping AI into existing systems risks metastasizing stagnation, using AI to redesign the systems themselves gives us a chance to remove, or at least reshape, both bottlenecks by combining computational scale with multiple competing theories of taste. »
« Second, we need to reward divergence… The opportunity isn’t randomness, but meaningful distance: things far enough beyond what we already know to expand our aperture, yet connected enough to give us a plausible way in. Instead of “people who liked X also liked Y,” AI can potentially identify an unexpected connection in aesthetic, rhythm, attitude, technique, or worldview between something unfamiliar and something we already value. »
« Third, we need to protect exploration. »
« Bring the three together, and we’re left with a very different idea of what a media platform could be. Not a single ranking logic, but many. Not optimization exclusively around demonstrated preference, but incentives for meaningful divergence. Not instant judgment by scale metrics, but protected space in which exploration and organic emergence can develop before exploitation takes over. »
