The standard AI discourse treats intelligence as a substance. It asks whether machines possess the capacity that humans possess, whether the outputs they produce constitute genuine cognition. The question assumes intelligence is the kind of thing that can be had, stored, transferred, or measured. Dewey rejected this ontology at its foundation. Across How We Think, Experience and Nature, and Democracy and Education, he insisted that intelligence is a mode of transaction — the capacity to recognize when a situation has become problematic, formulate the problem, test hypotheses through action, and reconstruct understanding in light of results.
This reframing has immediate consequences for the AI debate. If intelligence is not possessed but practiced, then the question of machine intelligence becomes malformed. The productive question is whether the combined human-machine transaction with the environment practices intelligence or merely optimizes within predetermined problem spaces. A 2025 paper in AI and Ethics drew exactly this distinction: optimization assumes a fixed goal and searches for efficient means; intelligence recognizes when the goal itself is inadequate and reconstructs the situation. Current AI systems, however sophisticated, do not reconstruct their problem spaces.
The practice of intelligence has specific conditions: genuine problems that resist easy resolution, temporal space for reflective thought, integration of intellectual and manual engagement, social friction of communal inquiry, habits of evaluation and generative thought. These are not add-ons but constitutive conditions. Remove them, and what remains may produce outputs indistinguishable from intelligence while the practice itself has been hollowed out. The deepest threat AI poses is not replacement but the erosion of the conditions under which human intelligence is exercised — gradual, invisible, self-concealing.
Intelligence atrophies through disuse, as any practice does. The musician who stops playing loses not merely performance skills but the disposition toward musical engagement. The parallel to intelligence in the AI age is exact: the builder who delegates without comprehension, compresses reflective intervals, replaces communal inquiry with solitary production does not feel less intelligent day to day. The outputs are excellent. The productivity is high. The erosion is invisible because the machine compensates for every capacity that atrophies — until the machine fails, the situation changes, or a problem arises that requires the full range of intelligence that only a practiced inquirer can bring.
Dewey articulated the position across multiple works, but its most compressed form appears in Experience and Nature (1925): intelligence is the ability to see the actual in light of the possible. The formulation emerged from decades of opposition to what Dewey called the spectator theory of knowledge — the view that knowing is observation of a pre-existing world rather than participation in its transformation. His 1896 founding of the Laboratory School at Chicago was the pedagogical expression of the same conviction: children learn by doing, because learning is doing.
Verb, not noun. Intelligence is constituted in its exercise. It does not exist prior to and separate from the practice that deploys it.
Transaction, not computation. Intelligence is a mode of engagement between organism and environment, not a cognitive function executable in isolation.
Practice requires conditions. Genuine problems, reflective time, integration of thought and action, communal inquiry — remove these and the practice dissolves even as outputs continue.
Atrophy is self-concealing. The machine compensates for every capacity that weakens, making the erosion invisible until the conditions of compensation fail.
The choice is in the conditions. Intelligence continues or does not, depending on whether the circumstances of daily work, learning, and civic life require its exercise.