CONCEPT
The Durkheim Test
Susan Leigh Star's 1989 proposal to evaluate AI systems by their capacity to
serve a community rather than by their ability to imitate an individual mind — an inversion of the
Turing Test that three decades of AI discourse largely ignored and the AI transition has made urgent.
The
Turing Test asks whether a machine can mimic individual human intelligence convincingly
enough to fool an interlocutor. The question is psychological: can this system pass as a person? Star's alternative asks a different question: does this system strengthen or weaken the social bonds of the community that uses it? Does it incorporate differing viewpoints or flatten them? Does it distribute capability or concentrate it? Does it create conditions for collective
deliberation or render deliberation unnecessary? The Durkheim Test is sociological rather than psychological, collective rather than individual, and its neglect for thirty-six years reflects the individualist assumptions embedded in
the culture that produced both the Turing Test and the technologies it was designed to evaluate. Applied to the AI tools of 2025 and beyond, the test produces concerning results:
democratization of capability is real, but the structural tendencies point toward weakened social bonds, flattened perspectives, and reduced