The subaltern is not merely the poor. Poverty is a material condition; subalternity is a structural one. The subaltern is the person whose experience of the social order is systematically different from the experience that the dominant narrative describes, and whose difference is rendered invisible not by suppression but by the narrative's claim to universality. The dominant narrative does not say "we are excluding your perspective." It says "we are describing reality" — and the description happens to coincide with the perspective of the dominant class while excluding perspectives that would challenge it. The AI transition has produced a global geography of subalternity that the dominant discourse maps as a geography of opportunity.
Gramsci used the term initially as a euphemism — a way of discussing class subordination in the Prison Notebooks without triggering the fascist censor's suspicion. But the concept outgrew its tactical origin. It became a tool for analyzing the specific condition of groups whose subordination consists not only in material deprivation but in epistemic exclusion — the exclusion from the processes through which a society produces its official account of reality.
Gayatri Spivak's 1988 essay "Can the Subaltern Speak?" extended Gramsci's analysis into postcolonial theory. The question is not whether the subaltern has thoughts — she does. The question is whether the institutions of the dominant order are structured to allow the subaltern's thought to achieve public circulation on its own terms, or whether the thought is always already mediated by the dominant class's categories, translated into the dominant class's frameworks, and thereby transformed into something that serves the dominant class's interests regardless of the subaltern's original intent.
The AI platform answers this question with disturbing clarity. The model was trained on data that overrepresents the English-speaking, technologically literate populations of the Global North. The model's embedded assumptions about what counts as knowledge, what counts as reasonable, what constitutes a good answer — these are the assumptions of this population, encoded into the model's parameters and presented as the neutral output of an objective computational process. The developer in Lagos who queries the model in English is not merely using a tool. She is absorbing a worldview, gradually and imperceptibly, through every interaction.
The Orange Pill invokes the developer in Lagos, the student in rural India, the entrepreneur in the Global South as evidence of democratization. The invocation is well-intentioned and partly accurate. But the developer in Lagos has access to the tool without access to the decisions that determine what the tool is, how it works, what it costs, and whose interests it serves. She is a user — empowered within the system, powerless over the system. The distinction between access and governance is the entire terrain of classical subaltern analysis.
Gramsci developed the concept across several notebooks, most explicitly in Notebook 25 on the margins of history, composed in 1934. He was drawing on Italian historiographical traditions but also responding to his own position as a prisoner whose perspective on Italian politics was being systematically excluded from the public record.
Subaltern Studies, a scholarly collective founded by Ranajit Guha in the early 1980s, extended Gramsci's framework to South Asian historiography, developing methodologies for recovering subaltern perspectives from archives constructed to exclude them. Gayatri Spivak's critical engagement with the project produced one of the most influential essays in postcolonial theory.
Structural, not merely material. Subalternity is a structural position of epistemic exclusion, not merely a condition of material deprivation.
Invisibility through universality. Dominant narratives render subaltern experience invisible not through suppression but through the claim to describe reality in universal terms.
Mediation of voice. Even when subaltern voices achieve circulation, they are typically mediated by dominant categories that transform their meaning before it reaches public discourse.
Access vs governance. The distinction between having access to tools and having voice in their governance is the defining structural feature of contemporary subalternity in the AI age.
Linguistic hegemony. The dominance of English in AI training data exports the epistemology of English-language culture as the model's general intelligence.
There is ongoing debate about whether the term "subaltern" retains analytical specificity or has been diluted through expansion — applied so broadly that it loses the structural precision Gramsci intended. Spivak's later work has emphasized that genuine subalternity involves irreducible opacity to the dominant discourse, not merely marginalization that could in principle be corrected through inclusion. The AI case raises new versions of this debate: can the subaltern speak through systems whose architecture encodes hegemonic values, or does the system's architecture itself require transformation before authentic subaltern voice becomes possible?