Muddling Through

In the AI Story

The argument proceeds from an epistemological observation. Comprehensive rational analysis — what Lindblom called the root method — assumes that an analyst can identify all relevant values, enumerate all possible alternatives, trace all consequences through the system, and select the optimum. Nobody has ever done this for any problem of genuine complexity. The information is inaccessible, the values are in irreconcilable tension, and the system's behavior is shaped by the responses of millions of actors whose decisions cannot be predicted in advance.

What actually works is the branch method: begin at the current situation, compare a limited number of alternatives that differ incrementally from the status quo, evaluate them against their marginal differences, and choose the increment whose practical consequences are most acceptable. The method accommodates disagreement about values without requiring resolution, because the alternatives are evaluated against consequences rather than against contested abstractions.

Muddling through has specific properties that make it effective in conditions of uncertainty: it preserves optionality, generates information, accommodates disagreement, and compounds. Each increment is an experiment whose consequences reveal features of the system that prior analysis could not have identified. The accumulated learning from thousands of small experiments builds a body of practical knowledge richer than any theoretical framework, because the knowledge describes what actually happens rather than what a model predicts should happen.

Applied to the AI transition, muddling through predicts — and the empirical record confirms — that effective governance is emerging not from any comprehensive national AI strategy but from thousands of incremental interventions: sector-specific regulations, organizational AI use policies, professional standards developed within communities of practice, and the daily adjustments of teachers, parents, and workers. The Berkeley study's AI Practice framework is muddling through at its best: modest, testable, revisable, grounded in eight months of observation inside a functioning organization.

Origin

Lindblom's 1959 article in Public Administration Review — 'The Science of Muddling Through' — appeared in a field dominated by the rational-comprehensive ideal inherited from Herbert Simon's bounded rationality and from the operations-research optimism of the postwar decade. The article was controversial immediately and remains the most cited article in the history of public administration.

Lindblom extended the argument through A Strategy of Decision (1963, with David Braybrooke) and Politics, Economics, and Welfare (1953, with Robert Dahl), developing what he eventually called disjointed incrementalism — the full theoretical apparatus behind the deliberately modest phrase.

Key Ideas

Root versus branch. The root method starts from fundamental values and derives optimal policy; the branch method starts from the current situation and compares incremental alternatives. Only the branch method has ever actually been practiced successfully for complex problems.

Analytical humility. The cognitive demands of comprehensive analysis exceed human capacity not occasionally but structurally. No amount of additional data or analytical sophistication closes the gap.

Values emerge through choice. Comprehensive analysis assumes values can be specified in advance; muddling through recognizes that values are clarified through the practical process of choosing between concrete alternatives.

Information generation. Each incremental intervention is an experiment. The accumulated data from thousands of experiments produces knowledge about how the system actually responds — knowledge no model could generate.

Failure as design. The first iteration of any AI governance intervention will be wrong. The question is not how to get it right but how to make its failure informative for the next iteration.

Debates & Critiques

The sharpest challenge to muddling through is gradual disempowerment — the argument that the method's locally rational steps can cumulatively erode the conditions for future democratic agency. The response is structural vigilance: extending the method to evaluate not only immediate consequences but effects on the system's capacity to take different steps in the future. The method must protect itself.