Monocultures of the Mind

In the AI Story

The framework connects Haraway's decades-long critique of the god trick to the specific pathologies of AI-generated content. The machine trained on the internet's dominant patterns produces outputs that gravitate toward the statistical center. The outputs are not neutral. They reflect the specific cultural, linguistic, and ideological composition of the training data — but they are received as neutral, because the fluent presentation performs universality. Users who think regularly with such a system are pulled toward the center along with the outputs. The fishbowl contracts around a specific set of culturally dominant shapes.

The agricultural metaphor is exact. A monoculture field produces high yields of a single crop. A biotic community produces a lower yield of any individual crop but sustains the soil, the water table, the surrounding species, the capacity to recover from disturbance. When agricultural yields became the single metric of success, biodiversity was lost — and with it the resilience of the whole system. Soil depletion followed. Pests adapted to the uniformity. Single-point-of-failure vulnerabilities emerged that the diverse system had not had.

Applied to cognition, the analogy illuminates a danger that neither the triumphalist nor the Luddite framings can see. The cognitive diversity of unaided human thought — the wildness of actual minds thinking in actual bodies in actual places — is what the AI system systematically flattens. Not intentionally. Structurally. The model is trained to produce the probable continuation. The probable continuation is the culturally dominant one. The user is trained, through exposure, to recognize the probable continuation as good.

The concept also connects to model collapse — the empirical phenomenon in which AI outputs trained on AI outputs degrade toward homogeneity. The cognitive version of model collapse is the one Haraway names: the progressive narrowing of what humans are able to think, not because anyone forbids the unusual thought but because the infrastructure of thinking has been optimized for the probable one.

Origin

Haraway has used the agricultural metaphor of monoculture throughout her work, drawing on her training in biology and her lifelong engagement with food politics. The specific application to AI appears in her 2026 interview with Laura Flanders, though the critique of cognitive homogenization as the deepest political danger of aggregation-based technologies has been present in her thought since the 1990s.

Key Ideas

Diversity is the substrate. The wild, situated, embodied diversity of human thought is what a cognitive biotic community requires.

Optimization is contraction. Optimizing for probable outputs contracts the space of what becomes thinkable.

The center is a specific somewhere. What counts as the probable continuation reflects specific cultural and ideological positions dressed as neutral.

Structural, not intentional. No one decides to produce monoculture. It emerges from the training dynamics of systems optimized for fluency.

Resilience requires biodiversity. A cognitive ecosystem that has lost its diversity has also lost its capacity to respond to the unexpected.