Arrested Development in Professional Practice

In the AI Story

The arrested-development finding is counterintuitive because it contradicts the common assumption that experience produces expertise. Ericsson showed that experience alone does not. What produces expertise is experience structured to meet the four conditions of deliberate practice. When those conditions are absent — when the work is routine, when feedback is infrequent or confounded, when the practitioner operates within the comfort zone rather than at the boundary — hours accumulate without development.

In medicine, the finding has been replicated in multiple studies. A 2005 systematic review in Annals of Internal Medicine by Choudhry and colleagues concluded that physicians who have been in practice longer may be at risk for providing lower-quality care — not because medicine got harder but because the practitioners' diagnostic models calcified around the patterns they had seen early in their careers and were not systematically updated through the disconfirming feedback that would have forced revision.

The AI era creates conditions for extending this pattern broadly. When tools handle the difficulty that professional work previously imposed, the incidental deliberate-practice that used to occur even without explicit design is eliminated. The developer who used to develop through debugging stops debugging. The lawyer who used to develop through careful case reading stops reading cases carefully. The physician who used to develop through diagnostic uncertainty loses the uncertainty. The professions that historically produced expertise as a byproduct of difficult practice will increasingly produce practitioners who look like veterans but have the representational architecture of novices.

The remedy, consistent with the broader framework, is the deliberate preservation of developmental conditions within AI-augmented workflows. This requires institutional recognition that experience and expertise are not the same thing, that the latter depends on specific conditions, and that those conditions must be maintained when the tools that previously imposed them have removed them.

Origin

The arrested-development finding emerged from Ericsson's studies of professional expertise across medicine, teaching, athletics, and management. It was systematized in his 2004 article Deliberate Practice and the Acquisition and Maintenance of Expert Performance in Medicine and Related Domains and extensively replicated in subsequent research.

Key Ideas

Experience ≠ expertise. Hours accumulated without the four conditions produce no representational growth.

Early calcification. Heuristics formed in training often solidify and are never revised, because routine practice provides confirmatory feedback that reinforces rather than challenges them.

Confirmation without correction. Routine cases provide feedback that confirms the practitioner's existing models even when the models are subtly wrong; only disconfirming feedback on novel cases forces revision.

AI amplification. Tool-handled difficulty eliminates the incidental deliberate practice that previously occurred even without explicit design.

Cross-professional pattern. The arrested-development finding generalizes across medicine, teaching, driving, judging, and other domains where practice conditions are routine rather than deliberate.