The Four-Year Gap
John Dewey died on June 1, 1952. Four years and two months later, at Dartmouth College in the summer of 1956, a group of mathematicians and engineers convened to launch the research program they called artificial intelligence. The gap between the philosopher of intelligence and the engineering of intelligence is four years. It is also a chasm. The two traditions that emerged — philosophical pragmatism with its participatory theory of knowledge, and computational cognitive science with its output-centered model of mind — have run in parallel for seven decades without fully encountering each other. The engineers built. The philosophers critiqued. Neither has reckoned with the other's central insight, and the AI transition now underway proceeds in the absence of the dialogue the gap has foreclosed.
The Productive Divergence — Contrarian ^ Opus
There is a parallel reading where the gap between Dewey's death and the Dartmouth Workshop represents not a missed opportunity but a necessary liberation. The engineering tradition needed to develop its own epistemic framework precisely because the philosophical tradition had reached its limits. Dewey's insistence on embodied experience and participatory knowledge would have strangled AI research in its crib, demanding that machines somehow replicate the full phenomenology of human consciousness before being credited with any form of intelligence. The gap allowed engineers to pursue what philosophy could not: the actual construction of systems that perform cognitive work, regardless of whether they experience it as humans do.
The supposed "haunting" by unasked questions—embodiment, experience, the felt difficulty of inquiry—may actually be the field's greatest strength. By bracketing these concerns, AI research discovered that intelligence-as-output can be mechanically instantiated without intelligence-as-experience. The frame problem and interpretability issues aren't distorted echoes of Deweyan questions but engineering challenges that emerge from actually building systems rather than theorizing about them. Consider the substrate independence that defines modern AI: neural networks running on silicon produce outputs indistinguishable from human reasoning in many domains. Had Dewey's framework dominated, researchers might still be debating whether a machine can truly "encounter" a problematic situation rather than building systems that solve problems. The gap didn't foreclose a dialogue; it enabled a research program that philosophy alone could never have initiated. The current AI transition proceeds not in the absence of philosophical insight but in the presence of engineering reality—a reality that demonstrates intelligence can be decomposed, formalized, and reconstructed without first solving the mystery of consciousness.
In the AI Story
The interval is not mere chronology. It is a conceptual separation. The Dartmouth organizers — John McCarthy, Marvin Minsky, Nathaniel Rochester, Claude Shannon — framed their research program around the assumption that intelligence could be defined by its outputs and replicated in any substrate capable of producing those outputs. This assumption would have struck Dewey as an instance of what he called the philosophical fallacy: the conversion of the outcomes of a process into the antecedent definition of the process itself.
Dewey had spent decades arguing that intelligence is not defined by its outputs but by its process. A machine that generates correct answers to mathematical problems is not, on Deweyan terms, doing mathematics. Mathematics, as a form of inquiry, involves the encounter with a problematic situation, the felt difficulty, the generation of hypotheses from experience, the testing against resistance, the reconstruction of understanding. The correct answer is the trace inquiry leaves behind, not the inquiry itself.
Four years earlier, this philosopher was alive and writing. Had the Dartmouth organizers consulted him — had the two traditions been in conversation — the entire trajectory of AI research might have been shaped by questions the engineers did not ask. What is the relationship between intelligence and embodiment? Between knowing and doing? Between the output of a process and the experience that produced it? These are not questions engineers typically found useful, but they are the questions whose neglect has now returned to haunt the AI transition. The hard problem, the frame problem, the interpretability problem — each is a distorted echo of a question Dewey's framework had already addressed.
The gap has consequences for the current moment. The dominant AI discourse treats intelligence as a substance to be engineered, measured, and deployed. The alternative tradition that would treat intelligence as a practice to be cultivated, measured by growth rather than output, has been largely absent from the conversation. The Dewey volume is an attempt to close the gap retrospectively — to apply the philosopher's framework to the engineering reality he did not live to see, and to ask whether the questions he asked seventy years ago still have purchase on the situation we face now.
Origin
The four-year interval is a historical fact. The conceptual gap it opens is a reconstruction by the Dewey volume, which uses the chronology as an entry point into the question of what philosophical pragmatism might have contributed to AI research had the two traditions been in contact at the founding moment. The framing is indebted to Hubert Dreyfus's extended critique of symbolic AI, which drew on Heidegger and Merleau-Ponty — parallel phenomenological traditions — but did not systematically engage with Dewey's pragmatism.
Key Ideas
Chronological proximity, conceptual distance. The philosopher and the engineering program were separated by fifty months, but their frameworks have never genuinely met.
Output vs. process. The Dartmouth framing defined intelligence by outputs; Dewey defined it by the process that produces them.
The philosophical fallacy. Defining a process by its outcomes converts the conclusion into the antecedent; the error underlies much of the AI discourse.
Retrospective dialogue. The Dewey volume is an attempt to stage the conversation that did not happen in 1952-56, applying the philosopher's framework to engineering reality he did not see.
Appears in the Orange Pill Cycle
The Necessary Incompleteness — Arbitrator ^ Opus
The right frame for understanding the four-year gap depends entirely on which question we're asking about intelligence. If we're asking "Can useful cognitive work be automated?"—the engineering question—then the contrarian view dominates (80/20). The Dartmouth approach enabled seventy years of concrete progress that Dewey's framework would likely have impeded. The output-focused definition of intelligence, however reductive philosophically, proved generative for building actual systems. But if we're asking "What is intelligence?"—the philosophical question—Edo's framing carries more weight (70/30). The neglect of process, embodiment, and experience has indeed returned as fundamental problems in AI alignment and interpretability.
The weighting shifts again when we consider practical consequences. For the development of AI capabilities, the gap was productive (90/10 contrarian). Engineering needed its own epistemology. But for understanding AI's social integration and human meaning, the missing dialogue matters enormously (75/25 Edo). The absence of process-oriented thinking has left us with systems we can build but not fully comprehend, outputs we can measure but not meaningfully evaluate. The frame problem isn't just an engineering challenge but a symptom of defining intelligence without reference to the experiencing subject.
The synthetic insight is that both traditions describe different aspects of a phenomenon too complex for either framework alone. The gap represents not a mistake but an inevitable specialization—philosophy attending to intelligence-as-lived, engineering to intelligence-as-performed. The current moment doesn't require closing the gap so much as recognizing it as constitutive: intelligence exists both as embodied process (Dewey) and as substrate-independent function (Dartmouth). The tension between these views isn't a problem to solve but the productive contradiction that drives both philosophical inquiry and engineering innovation forward. The four-year gap is less a chasm to bridge than a generative distance to maintain.
Further reading
- Hubert Dreyfus, What Computers Still Can't Do (1992).
- Pamela McCorduck, Machines Who Think (1979; revised 2004) — history of AI including the Dartmouth Workshop.
- Nils Nilsson, The Quest for Artificial Intelligence (2010).