Process Over Output

For most of the history of knowledge production, output was a reliable proxy for process. The essay that demonstrated understanding of a subject had been produced through the process of engaging with that subject; the professional brief that showed mastery of case law had been produced through the process of studying case law. This correlation was not perfect—plagiarism is as old as writing—but it was reliable enough that the institutions built to evaluate competence could proceed on the assumption that the output they were assessing evidenced the process that supposedly produced it. That assumption is now false, and its falseness is the deepest institutional challenge the AI transition presents. Clifford Geertz’s insistence on the primacy of process over output—the argument that the same behavior can mean fundamentally different things depending on the process that produced it, that a wink and a twitch are identical at the level of observable output but opposite in meaning—is the methodological