Donald Schon

In the [YOU] on AI Field Guide

Schon’s framework arrives in the cycle as the most precise account available of what changes when a builder describes a problem in natural language and receives an implementation back. In the prior era of software development, the workflow was organized according to the principles of technical rationality with a fidelity Schon would have recognized instantly: specification (the pre-defined problem), documentation study (the body of applied science), implementation (the application of technique), testing against specification (verification). The flow was linear, downward, from theory through technique to application. The language interface broke the linearity. When a developer describes a problem to Claude and receives an implementation, evaluates it, adjusts her description, and receives a revised implementation—in a continuous cycle—the workflow is no longer linear. The problem is no longer given. The developer and the tool are engaged in something that looks, structurally, like the conversation with the situation that Schon documented in every domain of expert practice.

The laparoscopic surgery example from the cycle’s account of collaboration is the paradigm case in Schon’s framework. The author was stuck on the Han argument, unable to find the pivot from acknowledging loss to showing what replaces it. He described the impasse to Claude. Claude’s response—the observation that surgeons who lost the tactile friction of open surgery gained the ability to perform operations impossible with open hands—was not a solution within the author’s frame. It was a reframing: the suggestion that friction removal at one level produces friction elevation to a higher level. In Schon’s terms, Claude provided back-talk that triggered a reframing. The practitioner had been operating within one frame; the machine’s response proposed an alternative. Reframing is the highest-order reflective skill Schon identified—the operation that distinguishes the master practitioner from the competent technician. The tool was performing it.

But the cycle also documents, with equal honesty, the danger Schon’s framework precisely identifies. The back-talk of the language interface is fluent regardless of its accuracy, structured regardless of its appropriateness, confident regardless of its justification. The practitioner must evaluate not just whether the response is correct but whether the response’s polish is concealing a fundamental misunderstanding. When the back-talk comes from clay or code or a patient’s body, the medium’s own logic grounds the evaluation. When the back-talk comes from a language model, the evaluation is harder: the response is smooth whether it is right or wrong, and the practitioner must develop what Schon called the practitioner’s repertoire—the judgment that distinguishes the genuine reframing from the plausible but empty suggestion—faster than any previous generation of practitioners had to.

The temporal dimension of reflection matters most here. Schon identified that reframing operates on a different timescale than iteration. The language interface accelerates iteration without accelerating reframing. The practitioner cycles through twenty move-backtalk-evaluate iterations in an hour while her capacity to reframe operates at the speed of cognitive reorganization. The result Schon’s framework predicts: rapid refinement within a fixed frame. The solution becomes increasingly polished while the frame itself—the definition of the problem, the assumptions structuring the approach—remains unexamined. Output accumulates. Understanding does not.

Origin

Donald Schon (1930–1997) was trained as a philosopher at Harvard, where he wrote a dissertation on the concept of displacement in John Dewey. He spent his early career in management consulting and urban planning before joining MIT, where he developed the framework of reflective practice through studies of professional education and practice across multiple fields. His 1983 book The Reflective Practitioner synthesized the framework across architecture, psychotherapy, engineering, town planning, and management.

The central adversary in Schon’s intellectual project was Herbert Simon, whose Sciences of the Artificial (1969) framed all intelligent behavior as search through a problem space. Simon’s model powered both classical AI and professional education: intelligence is optimization, expertise is efficient search, the professional applies formal knowledge to pre-specified problems. Schon’s observation was that the best practitioners—the ones who navigated the swampy lowlands with genuine competence—were not doing what Simon described. They were constructing the problem space rather than searching within it. The 1983 book demonstrated this through close analysis of recorded design conversations, clinical sessions, and management meetings. The 1987 follow-up, Educating the Reflective Practitioner, argued that professional schools had been designed to produce the wrong kind of practitioner and proposed a redesign around what he called the reflective practicum.

Schon’s irony—which he did not live to articulate—is that the machines built on Simon’s epistemology proved his framework in practice. Large language models do not search a problem space defined by logical operators. They generate responses through a process closer to the pattern-recognition-across-vast-implicit-knowledge that resists the clean hierarchy of basic science, applied science, and practice. And the tool they have become, in practitioners’ hands, converts the linear logic of technical rationality into the iterative logic of reflective practice. Simon and Schon debated for decades about the nature of professional knowledge. The machines built on Simon’s model have, in their practical effect, proved Schon right.

Key Ideas

Technical rationality and its failure. The dominant epistemology of professional education holds that practice is the application of scientific theory to practical problems. It works on the high ground of professional work—well-defined problems with established techniques. It fails in the swampy lowlands where professional competence actually matters most, because it assumes the problem is given when in fact the hardest professional work is figuring out what problem you are solving. Problem setting is the operation that technical rationality cannot account for, and it is the operation that distinguishes the master practitioner from the competent technician.

Reflection-in-action. The specific cognitive operation in which thinking and doing fuse—the practitioner adjusts her actions based on what she is observing about the results of her actions, in real time, without leaving the situation to consult a textbook. The surgeon who adjusts mid-operation, the architect whose sketch talks back to reveal an implication she had not intended, the jazz musician who hears the ensemble’s response and redirects the solo before the phrase ends: all are practicing reflection-in-action. The competence cannot be reduced to the application of prior knowledge. It requires the constant construction of new knowledge through the conversation with the situation.

The conversation with the situation. Schon’s name for the three-move cycle that constitutes all expert practice: propose, listen to the back-talk, evaluate and adjust. The sketch reveals what the architect did not intend. The patient’s body responds in ways the treatment plan did not anticipate. The conversation is genuine: both parties contribute something the other did not provide. The language interface has given practitioners a conversational partner of unprecedented breadth whose back-talk can trigger reframings across domains that no single human repertoire could traverse.

The practitioner’s repertoire. Reflection-in-action is not improvisation. It draws on an accumulated body of examples, patterns, and moves that the practitioner has internalized through years of engagement with situations that resist. The repertoire is what allows rapid recognition—the sense that “this is like that case from three years ago”—and what provides the evaluative capacity to distinguish back-talk that reveals a genuine dimension of the problem from back-talk that merely echoes the practitioner’s assumptions in a more polished form. When the back-talk from Claude is richer than the practitioner’s repertoire can evaluate, the conversation continues without the corrective function that makes it productive.

The asymmetric collaboration. Claude satisfies two of the three conditions for a reflective partner: it brings a different repertoire, and it makes its contributions in a form the practitioner can engage with. It fails the third: it does not reflect. The practitioner reflects and changes; the machine processes and responds. The collaboration produces learning in one direction only. This one-directional asymmetry places the full weight of evaluation on the practitioner—demanding that she perform both her own reflection and the compensatory evaluation for the partner’s failure to reflect.