Imperialist White Supremacist Capitalist Patriarchy

In the AI Story

The phrase emerged from hooks's frustration with progressive movements that would challenge one form of domination while leaving others unexamined. She watched white feminists challenge patriarchy while benefiting from and reproducing white supremacy. She watched socialist movements challenge capitalism while maintaining patriarchal structures. She watched anti-racist activists challenge white supremacy while ignoring the ways capitalism and imperialism shaped racial formation. Each movement's partial vision allowed it to claim radicalism while leaving the interlocking structure largely intact. hooks's insistence on naming the whole system at once was a refusal of this fragmentation. It forced every analysis to reckon with complexity, to acknowledge that no liberation movement that leaves any pillar of the structure standing has achieved liberation.

Applied to AI, the framework reveals what single-axis analyses miss. AI companies (capitalist structures) are concentrated in the Global North (imperialist geography), deploy products trained predominantly on English-language texts produced by white Western culture (white supremacist training data), and are led overwhelmingly by men (patriarchal leadership). The structure reproduces itself at every level: in who gets funded, whose knowledge is encoded, whose perspective the models center, whose labor is exploited in the supply chain, and whose communities bear the environmental costs of computational infrastructure. Each pillar supports the others. Addressing 'bias' in AI without addressing the capitalist incentive structure that rewards extraction over equity, the imperialist geography that concentrates capability in former colonial centers, and the patriarchal culture that shapes whose voices are heard in the industry is to mistake a symptom for the disease.

Researchers working at the intersection of AI and social justice have increasingly adopted hooks's integrated framework. The PMC publication on Black feminism and AI cites hooks alongside Patricia Hill Collins, Kimberlé Crenshaw, and Audre Lorde to analyze how 'AI-based sorting, classification, and prediction reproduce the specific dynamics of racial and gender domination.' The analysis treats AI not as a neutral technology with unfortunate biases but as a system produced by and reproducing the interlocking structures hooks named. The training data is not accidentally skewed; it is structured by centuries of decisions about whose knowledge gets written, published, preserved, and digitized—decisions shaped at every step by the systems of domination hooks spent her life making visible.

Origin

hooks began using variations of the phrase in the early 1980s, and by the 1990s it had become her signature diagnostic term. The phrase built on Black feminist theory's concept of intersectionality, Crenshaw's framework for understanding how race and gender (and class, and sexuality) produce unique forms of oppression that cannot be understood by adding separate analyses together. But where Crenshaw's framework was primarily analytical—a tool for legal and social analysis—hooks's phrase was both analytical and rhetorical. The awkwardness was the point. Saying the whole phrase aloud is uncomfortable. The discomfort is pedagogical. It forces the speaker and the listener to hold the whole structure in view, to resist the temptation to simplify, to sit with the complexity that genuine understanding requires.

Key Ideas

Interlocking systems. The four structures are not separable; white supremacy operates through capitalist mechanisms, patriarchy is funded by imperialist extraction, each system requires the others to function.

AI as structural reproduction. The technology industry's concentration in the Global North, its extraction of training data from global knowledge commons, its leadership demographics, and its profit model all reproduce the interlocking systems hooks named.

Partial liberation is no liberation. Movements that challenge one pillar while leaving others intact do not threaten the structure; they reorganize it, producing a more sophisticated and therefore more durable form of domination.

Training data encodes domination. The corpora on which large language models are trained are not neutral collections but specific, historically produced archives shaped at every stage by decisions about whose knowledge counts—decisions made within and by the structures hooks named.

Wholeness as method. The insistence on naming all four systems simultaneously is not rhetorical excess but analytical necessity; only by holding the whole can the interdependencies be perceived.