Reflection-in-Action

In the AI Story

The standard model of professional knowledge — technical rationality — assumes thinking and doing are sequential. First you think, then you do. First you diagnose, then you treat. Reflection-in-action violates this sequence constantly, not because practitioners are sloppy but because professional situations are too particular, too resistant to pre-formulation, for any plan to survive contact with reality. The plan is a hypothesis. The execution is the experiment. The competent practitioner treats it as such — observing divergence from expectation, adjusting the hypothesis, testing the adjustment through modified action.

The phenomenon is not trial and error. Trial and error is random. Reflection-in-action is structured by the practitioner's repertoire — the accumulated body of patterns, moves, and frames from past practice. The aberrant anatomy does not produce panic; it produces recognition. This is like that case from three years ago. This fibrosis means I need to change my approach. The recognition is instantaneous, often pre-verbal, and deeply informed by thousands of previous encounters that have deposited layers of perceptual competence too dense to articulate. Schon borrowed tacit knowledge from Michael Polanyi to name what the recognition draws on.

The arrival of large language models has transformed the conditions under which reflection-in-action occurs. Previous tools offered limited back-talk. The compiler said yes or no. The test suite said pass or fail. These responses were precise but thin. When a practitioner describes a problem to Claude and receives an implementation, the response is substantive enough to trigger genuine reflective work — it reveals dimensions of the problem the practitioner had not articulated. This is back-talk of a new order, and it enables reflection-in-action in domains where practitioners previously lacked the repertoire to sustain it.

The risk is calibrated to the gain. Reflection-in-action depends on the practitioner's capacity to evaluate the back-talk — to distinguish productive surprise from misleading pattern-match. When the back-talk comes from the clay, the patient, the code under test, the evaluation is grounded in the medium's own honest response. When the back-talk comes from a language model, the response is fluent regardless of accuracy, structured regardless of appropriateness, confident regardless of justification. The practitioner must evaluate whether the polish is concealing a seam — whether the smooth surface is hiding the place where the reasoning breaks. The ladder of inference runs faster than the evaluative muscle can follow, and what looks like reflection-in-action can collapse into iteration without evaluation — the external form of Schon's concept wearing the costume of reflective engagement.

Origin

Schon developed reflection-in-action in the late 1970s through field studies of professional practice — design studios, psychotherapy sessions, engineering offices. The concept crystallized in The Reflective Practitioner (1983) around the analysis of a single studio interaction between a master teacher he called Quist and a student named Petra, in which Quist's reframing of a stuck design problem exemplified the structure of move, back-talk, evaluation, and adjustment that the concept names.

Key Ideas

Thinking fused with action. Not retrospective analysis, not prospective planning — cognition inseparable from performance.

Structured by repertoire. The response to surprise is informed by thousands of previous encounters, not assembled from rules.

The situation talks back. The medium responds in ways the practitioner did not predict, and the response is the stimulus for the next move.

Evaluation is the hinge. Reflection-in-action depends on judging the back-talk accurately — distinguishing productive surprise from pattern-matched plausibility.

AI raises the stakes. The richer the tool's back-talk, the more demanding the evaluative function becomes, and the more often the fluency trap substitutes iteration for reflection.

Debates & Critiques

The most sustained critique of reflection-in-action comes from cognitive psychologists who argue Schon conflates distinct mental operations — conscious deliberation, tacit pattern recognition, and automatic motor response — under a single phenomenological banner. Defenders respond that the conflation is the point: at the level of phenomenology, the operations are fused in skilled performance, and disaggregating them analytically loses what makes expert practice expert. The AI moment sharpens the question: when the tool's response arrives in seconds, is the practitioner's evaluation genuine reflection or a rapid, pattern-matched assessment that wears reflection's clothing without its substance?