The Practitioner's Repertoire

In the AI Story

The repertoire has properties that matter enormously in the context of AI. First, it is built through friction. The diagnostician's repertoire was not assembled from a textbook. It was assembled from the ten thousand X-rays she looked at wrong before she started looking at them right. Each error forced attention. Each surprise disrupted expectation. The disruption created the conditions for reorganization of the existing structure. Without friction, the layers do not form properly — the deposition is superficial, the integration incomplete. This is the argument The Orange Pill makes in its discussion of ascending friction: removing friction at one level risks eliminating the conditions under which repertoires form at all.

Second, the repertoire is tacit. The diagnostician cannot articulate the rules she uses to read the X-ray because her competence does not consist of rules — it consists of perceptual patterns too complex and context-dependent to be captured in propositional form. Ask her how she saw the mass and she will say, "It just didn't look right." The description conveys nothing to the novice but is, in fact, a precise report of a perceptual event. Something in the image triggered a pattern her repertoire recognized as deviant, and the recognition was immediate, pre-verbal, and certain in a way no explicit reasoning could have produced. This is tacit knowledge operationalized as professional perception.

Third, the repertoire is personal. Two diagnosticians with thirty years of experience will have different repertoires, because they have seen different patients, made different errors, worked in different contexts. The repertoire is indexed by lived experience — the specific cases that taught the practitioner something she did not know. This personal indexing is what makes the repertoire irreducible to external representation: it cannot be transferred, copied, or uploaded, because it is organized by a life only one person has lived.

The computational repertoire that AI brings to the conversation is different in kind. It is not organized by lived experience but by statistical co-occurrence — patterns that appear together frequently are associated more strongly. The organization is powerful, producing connections the practitioner had not considered and drawing on domains she has never entered. But the organization is different: the practitioner's repertoire is indexed by significance (what mattered to her), while the computational repertoire is indexed by frequency (what appeared together often). These are not the same thing. The case that taught the diagnostician to look twice at the mediastinal silhouette is significant not because it was common but because it was consequential. The computational repertoire may miss it entirely, or surface it alongside a thousand statistically similar but clinically different patterns the diagnostician must somehow evaluate.

Origin

Schon developed the repertoire concept across The Reflective Practitioner (1983) and Educating the Reflective Practitioner (1987), building on Polanyi's tacit dimension and on his empirical observations of how master practitioners recognize situations and respond to them without recourse to explicit rules.

Key Ideas

Perception-organizing, not rule-containing. The repertoire shapes what the practitioner sees, not what she deduces.

Built through friction. Errors, surprises, and difficulties are the mechanism of deposition; without them, layers do not form.

Indexed by significance. The repertoire is organized by what mattered in the practitioner's lived experience, not by statistical frequency.

Irreducibly personal. Two experts in the same field have different repertoires because they have lived different practices.

Complementary to AI's computational repertoire. The machine's frequency-organized breadth and the practitioner's significance-organized depth can combine productively — if the practitioner retains evaluative primacy.

Debates & Critiques

Defenders of formal knowledge argue that the repertoire is ultimately reducible to cases plus rules for retrieving them, and that sufficiently sophisticated case-based reasoning systems could replicate what Schon attributes to tacit competence. Schon's defenders respond that the reduction misses what makes the repertoire work: the felt significance that weights patterns by their consequential importance in the practitioner's life, which no dataset can reconstruct because the significance is a function of stakes the dataset does not have.