Cultural Imperialism
Cultural imperialism occurs when a dominant group's experience, values, and cultural products are established as the universal norm, and other groups are marked as deviant, inferior, or invisible. The dominant group does not need to actively suppress other cultures; it simply occupies the default position — the unmarked category against which all others are measured. The English speaker does not know she speaks a particular language; she speaks 'language.' The viewer of Western oil painting does not know she is viewing a tradition; she is viewing 'art.' The normalization is so complete that it becomes invisible to those who share the normalized perspective.
The Infrastructure Determines the Universal — Contrarian ^ Opus
There is a parallel reading that begins not with cultural content but with material substrate. The 'universality' of English in AI systems reflects the fact that English-speaking institutions built the internet, encoded its protocols, and determined what would be digitized at scale. The dominance is not primarily cultural imperialism—a superstructure phenomenon—but infrastructure imperialism. The GPU clusters, the submarine cables, the content delivery networks, the venture capital that funds model training—these material conditions determine what becomes computationally tractable long before any questions of cultural values arise.
This reading suggests the remedy Young proposes—deliberative inclusion to restructure optimization criteria—mistakes effect for cause. You cannot deliberate your way out of infrastructural path dependence. Adding Yoruba speakers to the decision-making process does not change the fact that Yoruba has vastly less digitized text, that labeling Yoruba data costs more per token, that serving Yoruba queries from data centers increases latency, that monetizing Yoruba content produces lower returns. The economic gradient runs one direction. Treating this as a problem of cultural perspective ('unmarked defaults') rather than political economy ('the substrate costs money and someone must pay') produces remedies that address symptoms while leaving causes intact. The real question is not whose cultural logic governs optimization but who owns the infrastructure and how its construction costs are distributed.
In the AI Story
Young's face has become the most analytically productive for understanding AI cultural politics. Large language models trained predominantly on English text encode the assumptions embedded in that text as universal truths. Image generation systems trained predominantly on Western visual traditions produce outputs that default to Western compositional conventions. Music generation systems reproduce the harmonic structures of commercially successful Western music while treating maqam, West African polyrhythm, and Indian classical drone as exotic variations on an unmarked default.
The enforcement mechanism is statistical averaging, which makes the imperialism computationally automatic and morally diffuse. No engineer decided that Appalachian balladry is inferior to pop. No product manager decided that ukiyo-e is a 'style' rather than a tradition. The imperialism is enacted by the training data's composition and the optimization function's structure — mechanisms that present themselves as neutral and whose cultural specificity is invisible to those who share the dominant perspective.
Young insisted that the remedy is not merely diversification of training data but structural transformation of the decision-making processes that determine what systems are optimized for, what counts as quality, and whose criteria of excellence govern evaluation. Adding non-Western traditions as 'styles' applied to a Western compositional logic reproduces the imperialism at a deeper level. Genuine response requires treating alternative traditions as alternative logics with their own standards — which in turn requires the deliberative inclusion Young's communicative democracy demands.
Origin
The term predates Young in Marxist and postcolonial literature, but Young's distinctive contribution was to extract it from its context of state-level cultural policy and apply it to the internal cultural politics of nominally multicultural societies. Her analysis drew on feminist consciousness-raising (the discovery that 'the universal' was male), critical race theory (the discovery that 'the universal' was white), and postcolonial theory (the discovery that 'the universal' was European). She showed how the same structural mechanism operated across all three domains.
Key Ideas
The unmarked default. Dominant-group experience occupies the position of 'no position,' making its specificity invisible.
Enforcement through normalcy. No active suppression is required; the structural default does the work.
Fish in water. Those who share the dominant perspective cannot see the water they breathe.
Statistical imperialism in AI. Training-data composition enacts cultural imperialism automatically, without any agent intending it.
Remedy requires restructuring. Diversifying data without restructuring optimization logic produces surface diversity over durable imperialism.
Appears in the Orange Pill Cycle
Substrate Encodes Logic, Logic Requires Substrate — Arbitrator ^ Opus
The weighting shifts depending on which causal arrow you trace. On the question of how dominance is experienced and reproduced in day-to-day interaction with AI systems, Edo's framing is essentially correct (95%). Users encounter cultural imperialism as unmarked defaults—the English speaker who doesn't know she speaks a language, the system that treats maqam as exotic variation. The phenomenology is precisely as Young described: enforcement through normalcy, fish in water.
But on the question of why those defaults emerged and what mechanisms sustain them, the contrarian reading carries substantial weight (65%). Infrastructure determines cultural possibility more than cultural preference determines infrastructure. The dominance of English reflects material facts: who built the internet, what got digitized, where compute concentrates, how returns on investment flow. You cannot deliberate your way to parity when the substrate itself embeds massive asymmetries in data availability, labeling costs, and economic returns.
The synthesis: cultural imperialism and infrastructure imperialism are the same phenomenon viewed at different scales. Culture becomes infrastructure (English dominance makes Unicode default to Latin ranges, makes ASR systems optimize for English phonemes). Infrastructure becomes culture (the material fact that English has more training data makes 'quality' in language models encode English-specific patterns). Young's remedy—restructuring optimization criteria through inclusive deliberation—is necessary but not sufficient. It must be paired with what she herself recognized in her later work on structural injustice: interventions that address the material conditions producing the asymmetries. The question isn't cultural logic versus political economy. It's recognizing that cultural logic is sedimented in infrastructure, and infrastructure encodes cultural logic.
Further reading
- Iris Marion Young, Justice and the Politics of Difference (Princeton, 1990), chapter 2
- Arturo Escobar, Designs for the Pluriverse (Duke, 2018)
- Walter Mignolo, The Darker Side of Western Modernity (Duke, 2011)
- Kate Crawford, Atlas of AI (Yale, 2021)