The Laparoscopic Surgery Transition
The laparoscopic transition is Collins's most powerful analogy for understanding what happens when a technology changes the medium of expert practice. In 1987, French surgeon Philippe Mouret performed one of the first laparoscopic cholecystectomies, removing a gallbladder through small incisions using a camera and instruments rather than direct hand contact. Within a decade, the procedure had become the standard of care, displacing the open technique surgeons had practiced for a century. The transition produced a generational fracture: surgeons trained on open technique possessed one form of expertise, those trained on laparoscopic possessed another, and the relationship between them was not the simple progression that 'improvement' implies.
The Institution's Absorptive Capacity — Contrarian ^ Opus
There is a parallel reading that begins not with the expertise transformation itself but with the institutional container that made it survivable. The laparoscopic transition succeeded because surgery possessed what AI-driven software development fundamentally lacks: thick institutional walls that constrained the pace and scope of adoption.
Consider what bounded the surgical transition. Credentialing bodies controlled who could attempt the new procedure. Hospitals required proctored cases before granting privileges. Malpractice regimes created accountability feedback loops. Insurance reimbursement schedules shaped adoption curves. Most critically, the patient's body provided unambiguous failure signals — complications, mortality, recovery times — that no surgeon could explain away and no marketing could obscure. These structures didn't prevent elevated complication rates during the transition, but they prevented catastrophic institutional collapse. They gave the system time to develop new training pathways, to identify which elements of the old tacit knowledge still mattered, to build mentorship relationships in the new technique.
Software development has none of this. No credentialing requirement gates who can deploy AI-generated code into production systems. No institutional review boards slow adoption to match training capacity. Most fatally, software's failure modes are diffuse, delayed, and deniable in ways surgical complications never are. A codebase can degrade for years before collapse, and when it does, attribution is nearly impossible. The laparoscopic analogy flatters the AI transition by importing the stabilizing institutional infrastructure without acknowledging its absence in the target domain. What we're actually experiencing isn't a contained expertise transformation within robust institutions — it's the transformation happening faster than any institution can form around it.
In the AI Story
The analogy illuminates both what is gained and what is lost in technological transitions that transform the medium of practice. Open surgeons possessed richly somatic tacit knowledge: the hand could feel tissue, detect anomalies, register resistance in ways that informed surgical judgment. The laparoscope removed this tactile channel, replacing it with two-dimensional video. Critics within the surgical community argued this represented a loss of surgical judgment's foundation. They were right about the change. They were wrong about the consequence.
What actually happened is what Collins's framework predicts: surgeons trained on laparoscopic technique developed different but equally demanding polimorphic expertise. They learned to construct three-dimensional spatial models from two-dimensional images, to coordinate instruments whose visual and motor spaces did not align, to compensate for lost tactile feedback with heightened visual sensitivity. The expertise was transformed, not degraded. But — and this is Collins's crucial addition — the transformation was transmitted through the same social mechanisms as the original expertise: apprenticeship, mentorship, proctored cases, the specific human relationship of experienced practitioner guiding novice through the procedure.
The parallel to the AI transition is both striking and imperfect. The imperfection matters. The laparoscopic transition took place within existing surgical institutions that had robust apprenticeship structures; the new technique was absorbed by those structures and transmitted through them. The AI transition is taking place faster than equivalent apprenticeship structures for AI-directed work can form. The laparoscopic surgeon of 1997 had mentors who had mastered the new technique. The AI-directed developer of 2026 has few such mentors, because the practice is too new for a mentorship tradition to have formed.
Origin
Collins developed the analogy across multiple papers and in Tacit and Explicit Knowledge (2010). The example draws on extensive sociological literature on surgical training, including work by Atul Gawande and others on the specific demands of laparoscopic expertise formation.
Key Ideas
Transformation, not loss. The new expertise was as demanding as the old, but differently constituted.
Social transmission persisted. Despite the technological change, expertise still transferred through apprenticeship.
Temporal asymmetry. Robust training structures took a decade to mature; complication rates were elevated during the transition period.
Imperfect AI parallel. The AI transition lacks the institutional apprenticeship infrastructure that cushioned the surgical one.
Appears in the Orange Pill Cycle
Institutional Velocity as Hidden Variable — Arbitrator ^ Opus
The transformation-not-loss claim is 100% right at the level of individual expertise formation. Laparoscopic surgeons did develop genuine, demanding polimorphic capabilities that weren't degraded versions of open technique. The AI parallel holds here: developers working fluently with LLMs are developing real skills in prompt engineering, output evaluation, and architectural judgment that weren't required before. The expertise transformation is authentic in both cases.
But the institutional infrastructure question splits 80/20 toward the contrarian view, because it determines whether the expertise transformation can happen at human-viable timescales. The surgical transition's decade-long maturation wasn't a bug — it was the feature that allowed mentorship structures to form, training curricula to develop, and complication rates to decline before the technique became universal. Surgery's credentialing and accountability mechanisms created the temporal slack required for the social transmission of expertise to function. Software development is attempting the same transformation at 10x speed with 1/10th the institutional support structure. This doesn't mean the transformation can't happen, but it does mean we're in uncharted territory regarding how long elevated 'complication rates' (unmaintainable codebases, conceptual debt, degraded systematic understanding) will persist.
The right synthetic frame might be expertise transformation velocity: not whether AI enables genuine new capabilities (it does), but whether human learning and institutional adaptation can match the technology's deployment speed. The laparoscopic case succeeded because institutional friction kept transformation velocity within human learning capacity. The AI case is testing what happens when that constraint is removed. Collins's framework explains the expertise transformation; it doesn't explain what happens when transformation outpaces the social structures required to transmit it.
Further reading
- Harry Collins, Tacit and Explicit Knowledge (University of Chicago Press, 2010)
- Atul Gawande, Complications (Metropolitan Books, 2002)
- Harry Collins, 'What's wrong with relativism?' (1995)