Computers Don't Give a Damn

In the AI Story

Winograd's essay, published in Boston Review in December 2024, applied Haugeland's principle to the domains where AI's absence of caring creates consequential risk. In medicine, a system interpreting symptoms and recommending treatments does not know what it is like to be sick, to sit across from a frightened patient whose questions are not really about prognosis but about meaning. In law, a system drafting briefs does not know what justice is in the experiential sense—does not know what it feels like to be wronged, to seek redress, to stand before a tribunal. In education, a system generating essays does not care whether the student learns or merely produces output. For many purposes, the absence is inconsequential—pragmatic competence suffices. For the cases that matter most, the gap between competence and caring is everything.

The principle connects to Winograd's career-long distinction between capability and character. Large language models demonstrate that vast territories of practical competence do not require understanding—they can navigate open domains, interpret ambiguity, produce effective outputs across every field of human knowledge. What they cannot do is mean anything by producing those outputs. Meaning requires stakes. The system recommending a medical treatment does not care whether the patient lives or dies. The system designing a building does not care whether it becomes a home. The system writing code does not care whether the product serves or harms. This indifference is not a defect to be fixed through better training—it is a structural feature of systems whose 'existence' is not at risk in their outputs.

Origin

Haugeland, an American philosopher of mind and cognitive science (1945–2010), developed the principle across his career studying artificial intelligence, Heideggerian phenomenology, and the nature of intentionality. His collected essays, including 'The Intentionality All-Stars' and 'Semantic Engines,' argued that genuine understanding requires what he called 'being in a position to give a damn'—a capacity grounded in embodied existence with consequences. Winograd, who knew Haugeland's work, found in this principle the compressed expression of what his fifty-year trajectory from SHRDLU through Heideggerian critique had been demonstrating: the gap between processing and understanding is ultimately a gap between indifference and caring.

Key Ideas

Caring as ontological condition. Not an emotion to be engineered but a consequence of being a creature with stakes—whose existence depends on outcomes, who can be helped or harmed, who inhabits vulnerability.

The indifference of competence. Systems achieving extraordinary pragmatic capability across medicine, law, education, and design do not care whether their outputs serve or harm—statistical optimization is neutral regarding human flourishing.

Meaning requires stakes. What outputs mean to people affected by them—what a diagnosis means to a patient, what a verdict means to a plaintiff—is inaccessible to systems lacking experiential substrate for suffering.

The irreplaceable human contribution. When machines can produce every output, the human capacity to care about what those outputs are for becomes the only check on capability deployed without purpose.