Rittel & Webber — "Dilemmas in a General Theory of Planning" (1973)

1. Source

Horst W. J. Rittel and Melvin M. Webber, "Dilemmas in a General Theory of Planning," Policy Sciences 4, no. 2 (1973): 155–169. Elsevier Scientific Publishing Company. (Adapted from a paper presented at the Panel on Policy Sciences, AAAS, Boston, December 1969.)

PDF: bibliography/deep-reads/rittel-webber-dilemmas-general-theory-planning-1973.pdf

2. What this work is about

Rittel and Webber's central claim is that the scientific-engineering paradigm of problem-solving — which underlies professional planning and policy analysis — is structurally inapplicable to social problems. The paper is written in the aftermath of the 1960s systems-analysis movement: the optimism of PPBS, national goals commissions, social indicator programs, and the "modern-classical model" of planning as a cybernetic governing system. All of that, Rittel and Webber argue, failed not for want of skill or resources, but because the problems being addressed are categorically different from the problems that science and engineering know how to handle. Social and policy problems are wicked — a term they use to mean malignant, vicious, tricky, aggressive — as opposed to the tame (or benign) problems of mathematics, chess, chemistry, and engineering.

The argument proceeds in two main movements. First, Rittel and Webber enumerate ten properties that distinguish wicked problems from tame ones: wicked problems have no definitive formulation; no stopping rule; solutions that are good-or-bad rather than true-or-false; no immediate or ultimate test; they are one-shot operations; they have no exhaustively describable solution set; every one is essentially unique; every one is a symptom of another problem; multiple valid explanations exist for any discrepancy; and the planner has no right to be wrong (unlike the scientist, who is immunized by falsificationism from responsibility for failed hypotheses). Second, in a briefer section on "The Social Context," they argue that the wickedness of planning is compounded by growing social pluralism: there is no unitary public, no aggregate welfare function, and no overriding social ethic — which means that any "solution" enacted by planners is necessarily partisan, a political act disguised as technical judgment. Planning is politics; there is no escaping it.

3. Key vocabulary

Tame problem: A problem that can be definitively formulated, has a stopping rule, admits true-or-false solutions, and allows trial-and-error learning. Chess, mathematical equations, chemical structure identification are examples. Tame does not mean easy.

Wicked problem: A problem that resists definitive formulation, has no stopping rule, admits only good-or-bad solutions, forecloses trial-and-error, is essentially unique, and is embedded in a hierarchy of other problems. All substantive social policy problems.

Definitive formulation: A complete problem statement from which all needed information can be derived in advance of solving. Possible for tame problems; impossible for wicked ones because understanding the problem requires knowing the solution space.

Stopping rule: A criterion internal to the problem that signals when the solution has been found. Absent from wicked problems; planners stop for external reasons (money, time, patience).

Re-solution: Rittel and Webber's preferred term for what planners do — not "solution" but re-solution, iterative partial addressing. "Social problems are never solved. At best they are only re-solved — over and over again."

One-shot operation: The property that every implementation of a wicked-problem response is irreversible and consequential — you cannot build the freeway as an experiment. Traces cannot be undone.

Systems-approach of the first generation: Operations Research style — define problem, gather information, analyze, synthesize, solve. Works only after the problem has been tamed. Contrast with "second generation" approaches based on argumentative processes.

Argumentative process: The mode of planning Rittel and Webber gesture toward as appropriate for wicked problems — an iterative process in which images of problem and solution emerge together through incessant judgment and critical argument among participants.

Planner has no right to be wrong: The tenth property, and the sharpest. Scientists are protected by the norm of falsificationism — hypotheses are offered for refutation, and failed ones are not moral failures. Planners bear direct moral responsibility for consequences; there is no equivalent immunity.

World view: Property 9 makes explicit that the analyst's "world view" is the strongest determinant of how a discrepancy is explained and therefore of how it is resolved. Explanation is not neutral.

4. Analytical moves

A. Historical scaffold. The paper opens by positioning itself inside a specific historical defeat: the failure of the 1960s systems-analysis movement and the rise of anti-professional public protest. This is not throat-clearing; it does argumentative work. The defeat is the evidence that something structural is wrong with the professional model, not merely that the practitioners were incompetent.

B. The competence paradox. A striking rhetorical move at the outset: professionals are now being attacked precisely as they acquire genuine competence. The problem is therefore not incompetence but category error — the tools are real but applied to the wrong kind of problem.

C. The efficiency genealogy. The professional paradigm of planning is traced to 18th century physics and classical economics — the principle of least means, the guiding concept of operations research. This genealogy is important: the tame-problem model is not accidental but is historically rooted in exactly the era when problems were tame (or seemed so) — definable, consensual, technical.

D. The tame/wicked distinction as a claim about problem ontology. This is the paper's central philosophical move. The distinction is not about difficulty — tame problems can be extremely hard — but about structural properties of the problem itself. Wickedness is intrinsic to the domain, not a consequence of inadequate methods.

E. The self-referential trap (Property 1). The most philosophically sharp of the ten properties: the information needed to understand a wicked problem depends on one's idea for solving it. Understanding and resolution are concomitant. You cannot first understand, then solve. This is a genuine circularity, not a mere observation about complexity. It structurally forecloses the classic systems-analysis sequence.

F. Problem-formulation as solution-specification. From Property 1 follows a corollary Rittel and Webber make explicit: "The formulation of a wicked problem is the problem." Every specification of the problem is simultaneously a specification of a direction of treatment. This means there is no neutral problem statement — every formulation already encodes a politics.

G. One-shot irreversibility as epistemic trap. Property 5 closes the loop: if every solution attempt is irreversible and consequential, trial-and-error learning is foreclosed. Combined with Property 1 (can't understand before solving) and Property 4 (no ultimate test), the planner is epistemically trapped — required to act with commitment under irresolvable uncertainty.

H. The symptom hierarchy (Property 8) against incrementalism. Problems are symptoms of higher-level problems. This gives the paper a direct argument against Lindblom-style incrementalism: a marginal fix at a low level not only fails to address the higher problem but may make the higher problem harder to solve (by reinforcing organizational patterns, locking in information-controller power, etc.). Computerization example is vivid: it reduces operational cost but makes structural change harder.

I. The political dénouement. The Social Context section (Section IV) applies the wicked-problem argument directly to the social preconditions of planning. If society is pluralistic with no unitary welfare function, then any plan is necessarily partisan — not a technical finding but a political choice enacted by someone who "is also the player in a political game, seeking to promote his private vision of goodness over others'." Planning is a component of politics. There is no escaping it.

5. Curiosities and open questions

The taming operation. A protocol is precisely an attempt to tame a wicked problem — to impose a definitive formulation, a stopping criterion, and a true/false verdict on something that resists all three. What does this mean? That every protocol is a gamble: it bets that the problem is tame enough, or can be made sufficiently tame, that the costs of the wrong formulation are acceptable. The Formalization Ratchet (CL-001) might be exactly what happens when that bet is locked in against new evidence.

Pre-taming as invisible political act. Property 1 says formulation is solution-specification. If so, then the act of writing a protocol — choosing what to formalize, how to draw the boundaries — is already a political act, but one that often masquerades as administrative or technical. The protocol naturalizes a particular formulation. What's lost is the alternatives that were available before the formulation closed.

Operations Research becomes operational only after taming. Rittel and Webber make this explicit: OR methods "become operational, however, only after the most important decisions have been made, i.e. after the problem has already been tamed." This is eerie. Formalization doesn't solve the hard part; it hides the hard part, which was the taming itself. The protocol body (the rules, the code, the procedure) is OR-operational. The taming was political.

The stopping rule substitution. Planners stop "not for reasons inherent in the logic of the problem" but for external reasons — money, time, patience. When a protocol is adopted, does it substitute an artificial stopping rule for the absent natural one? The protocol says: this is the stopping criterion, this is the test of compliance. The protocol's stopping rule is a political decision posing as a logical criterion.

Property 9 and the explanation-choice problem. The choice of explanation for a discrepancy determines the resolution direction, and this choice is arbitrary in the logical sense — "attitudinal criteria guide the choice." Protocols encode one explanation (one causal story) and institutionalize it. When the explanation turns out to be wrong, the protocol persists anyway (CL-001). The discrepancy between the protocol's embedded explanation and the actual causal structure is never legible from inside the protocol.

The planner has no right to be wrong — does this describe protocols? Scientists have Popperian immunity: failed hypotheses are not moral failures. Planners do not. But protocol designers occupy an intermediate position: the protocol design is treated as technical (thus potentially wrong without moral blame), but the consequences of its failures fall on everyone subject to it. This asymmetry might be load-bearing in understanding why protocols resist revision: revision would require someone to accept moral responsibility for the period when the wrong protocol was in force.

The argumentative process as protocol-design methodology. Rittel and Webber gesture at "second generation" planning — argumentative, iterative, involving incessant judgment among participants. This sounds like how protocols should be developed, but the paper offers no mechanism. The structure of wicked problems seems to preclude any methodology. Is there a mode of protocol design that preserves the argumentative quality rather than foreclosing it? Or does the act of formalization necessarily terminate the argument?

The levels problem and formalization level-choice. Property 8: there is no natural level at which a wicked problem should be addressed. Higher = more general, harder to act on; lower = more tractable, but risks reinforcing the symptom and making the higher problem harder. Where a protocol is pitched determines what it can and cannot address. And where it is pitched is not determined by the problem itself — it is a choice, made on political and organizational grounds, and then frozen.

6. Research connections

CL-001 (Formalization Ratchet). The tame/wicked distinction illuminates why the ratchet exists. Formalizing a wicked problem requires imposing a stopping rule, a definitive formulation, and a true/false test — all of which are artificially imposed, not logically derived. Once imposed, the formulation naturalizes: the protocol and its embedded problem-formulation become mutually reinforcing. Revision requires re-opening the formulation, which means re-entering the wicked territory the protocol was designed to escape. The Formalization Ratchet is, on this reading, the cost of taming: you get out of wickedness by closing the argument, and closing the argument makes it hard to reopen.

CL-001 (internally developed vs. externally imposed distinction). Property 1's circularity suggests a hypothesis about this open question: internally developed protocols might encode a richer sampling of the problem's wicked properties (because the developers lived inside the problem space), while externally imposed protocols are more likely to encode a single politically convenient formulation. Both will ratchet, but internally developed ones may ratchet with more embedded problem-awareness — or the opposite, since insiders may also be most invested in a particular problem-framing.

CL-002 (Coordination Cost Conservation). The symptom hierarchy (Property 8) maps onto this. If solving a lower-level coordination problem (computerizing an administrative process) makes the higher-level coordination problem harder (structural change becomes more costly), then coordination cost is indeed displaced rather than eliminated — it is shifted upward to a less tractable level.

F-001 (Ossification). The ten properties together suggest that ossification is not a failure of institutional memory alone but a structural consequence of taming. Every protocol that successfully tames a wicked problem generates ossification as a byproduct, because the taming closes the formulation. The rate of ossification may track the degree of wickedness of the original problem — the more wicked the problem was before formalization, the more the protocol forecloses revision.

F-002 (Hardness Asymmetry). Property 1 gives a mechanism: formalization is easy because it substitutes an artificial problem formulation for the real one, which requires no understanding of the actual wicked problem. Deformalization is hard because it requires re-entering the wicked territory — reopening a formulation that has been naturalized. The asymmetry is built into the structure of wickedness.

F-003 (Goodhart). Property 9 is closely related. The explanation selected for the discrepancy determines the resolution; once the resolution is institutionalized (measured, targeted), the measure becomes the proxy for the explanation. When the explanation was wrong (and there was no way to know it was wrong), the measure optimizes for the wrong thing. Goodhart is the downstream consequence of Property 9's arbitrary explanation-choice, operating through time.

C items: C-003 (rules-as-code lowering boundary search cost) — the cost reduction happens in the tame part; the wicked part was already paid. C-004 (update opacity) — directly connected to Properties 1 and 9: the explanation embedded in the protocol is opaque because it was never stated as a hypothesis, only as a formulation. C-011 (notation lock-in) — Rittel and Webber's problem with notation: a mathematical formulation of a wicked problem is a notation that naturalizes one problem framing.

7. What it opens

Rittel's subsequent work. Rittel developed IBIS (Issue-Based Information System) as a methodology for argumentative planning — this is the "second generation" approach gestured at in the paper. Investigating IBIS would reveal whether there is a methodology that preserves wickedness rather than taming it.

Lindblom's incrementalism. Rittel and Webber make a direct argument against Lindblom. Reading Lindblom's original "Science of Muddling Through" (1959) against this paper would clarify whether the symptom hierarchy (Property 8) genuinely defeats incrementalism or whether Lindblom has a response.

Questions opened for Humboldt: - If a protocol is a taming device, what determines which formulation gets encoded? This is a political economy question about protocol design, not just protocol behavior. - Is there a meaningful distinction between protocols that were designed to tame wicked problems and protocols that emerged from already-tame domains (technical standards, file formats)? Do they ossify differently? - Property 10 ("planner has no right to be wrong") raises a question about accountability asymmetries in protocol governance: who bears moral responsibility for a wrong protocol, and how does the answer affect revision behavior?

8. What it says about becoming a better researcher

Rittel and Webber's most important lesson for research methodology is embedded in Property 1: the information needed to understand a problem depends on one's idea for solving it. This is a warning against premature problem-formulation. When approaching a new domain, the researcher who commits too early to a problem statement has already implicitly committed to a solution direction — and has closed off the wicked territory that might have generated better understanding. The patience to remain in the wicked phase — to resist formulating — is a research discipline.

The paper also implicitly criticizes the "systems-approach of the first generation" as a research style: understand, then solve. For wicked problems, understanding and resolution are concomitant. Research that seeks to achieve full understanding before generating candidate laws may be applying a tame-problem model to wicked-problem territory. The argumentative, iterative approach — where candidate formulations (hypotheses) are generated early and subjected to incessant critical argument — is the mode appropriate to wicked inquiry.

Finally, the tenth property ("planner has no right to be wrong") has a research analogue: the researcher who treats their candidate laws as published positions rather than hypotheses offered for refutation has given up Popperian immunity. The health of the research is partly the health of the researcher's willingness to be falsified. This is what F items are for.

Reading log

Read: 2026-06-13
Method: Full PDF read (11 pages), actual document, no substitution from training knowledge
C items generated: C-015 through C-022 (8 items)
Time in text: complete read of all 15 pages (155–169)