Disjointed Incrementalism
Disjointed incrementalism is the name Lindblom and David Braybrooke gave to the mature theoretical formulation of muddling through. The 'disjointed' modifier signals that the method does not operate through integrated, top-down coordination but through distributed, partial, and often uncoordinated interventions by many actors pursuing different objectives. The 'incrementalism' signals that each intervention is small and builds on existing arrangements. Together, they describe a mode of institutional intelligence that is messy, contested, perpetually revised, and — in conditions of complexity — more effective than the coordinated comprehensive alternative.
In the AI Story
Braybrooke and Lindblom's 1963 book A Strategy of Decision laid out eight specific features of disjointed incrementalism: margin-dependent choice, restricted consideration of alternatives, restricted consideration of consequences, adjustment of objectives to policies, reconstructive treatment of data, serial analysis and evaluation, remedial orientation, and social fragmentation of analysis. Each feature contrasts with a corresponding feature of the synoptic ideal.
The 'disjointed' character is essential. Disjointedness is not disorganization — it is the condition under which many analysts and decision-makers, operating with different perspectives and interests, can contribute their partial knowledge to collective decisions without any single authority integrating their contributions into a comprehensive plan. The integration happens through interaction, not through design.
Applied to AI governance, disjointed incrementalism describes what is actually happening: professional associations developing sector-specific standards, individual organizations crafting AI use policies, regulators issuing targeted interventions, courts adjudicating specific disputes, parents making household decisions. Each actor operates with partial knowledge, pursues their own objectives, and responds to the behavior of other actors. The aggregate of these interactions constitutes a de facto AI governance framework that no central authority designed — and that is, on Lindblom's analysis, superior to any framework a central authority could design.
The deliberate disjointedness is the point. A single, integrated AI governance framework would require agreement on values that democratic societies cannot achieve, information that analysts cannot gather, and coordination that institutions cannot maintain. Disjointed incrementalism acknowledges these constraints and proceeds anyway, extracting from the partial contributions of many actors more collective intelligence than any single authority could produce.
Origin
Braybrooke and Lindblom developed the full framework in A Strategy of Decision (1963), which expanded the 1959 article's informal treatment into a rigorous theoretical account. The book's subtitle — 'Policy Evaluation as a Social Process' — encoded the disjointed character: evaluation is social, distributed, and proceeds through interaction.
Key Ideas
Margin-dependent choice. Decisions are made at the margin of existing arrangements, not from a clean slate.
Restricted consideration. Both alternatives and consequences are examined selectively, focused on the differences that matter.
Adjustment of objectives. Goals are revised as constraints become clearer through the process of choosing among means.
Serial and remedial. Policy proceeds through a sequence of small adjustments oriented toward remedying specific observed problems rather than achieving specified positive goals.
Social fragmentation. Analysis is distributed across many actors with different perspectives, and the fragmentation is a feature that enlarges collective intelligence.
Appears in the Orange Pill Cycle
Further reading
- David Braybrooke and Charles Lindblom, A Strategy of Decision (1963)
- Charles Lindblom, The Intelligence of Democracy (1965)