Institutional Pluralism
In The You On AI Field Guide
The principle connects directly to Landes's argument that experimentation across jurisdictions was what made successful industrial adaptation possible. Different nations, different states, different municipalities trying different regulatory approaches, different educational models, different labor market interventions — and competitive pressure between jurisdictions ensuring that successful experiments are copied and unsuccessful ones are abandoned. The selection mechanism operates on experiments, not on theories.
The contemporary application bears directly on international AI governance debates. Proposals for unified global AI regulation assume that the right institutional response is known and can be imposed uniformly. Landes's framework suggests the opposite: the right institutional response is not known, cannot be known in advance, and must be discovered through the parallel operation of many approaches. Uniformity forecloses the selection mechanism on which adaptive institutional development depends.
The tension with safety concerns is real. Institutional pluralism can produce races to the bottom as jurisdictions compete to attract AI investment by relaxing safety standards. Landes's framework does not deny this risk but insists that the alternative — uniform standards imposed by authority that cannot learn from experimentation — produces worse long-run outcomes. The resolution lies in identifying which questions require uniformity (certain safety floors) and which require experimentation (most regulatory architecture), rather than assuming that one approach fits all.
Origin
The concept is developed across Landes's work on European industrialization, drawing on comparative analysis of how different national regulatory frameworks produced different adaptation trajectories during the Industrial Revolution.
Key Ideas
Distributed authority as innovation substrate. The structural condition under which good ideas can survive suppression by finding alternative harbors.
Competition between jurisdictions as selection mechanism. Successful institutional experiments are copied; unsuccessful ones are abandoned; the system learns without any central authority doing the learning.
Uniformity forecloses learning. Single approaches, imposed uniformly, cannot self-correct because they lack the comparative evidence selection requires.
The AI application. National and international AI governance should preserve institutional pluralism precisely where the right approach is unknown.