Skip to content

Policy Design

The structured activity of crafting an authorised rule for a target population of intentional agents inside an institution, factored into six separable components — frame, goal, target, instrument, implementation, evaluation — so failure can be localized to a slot rather than pronounced globally.

Core Idea

Policy design is the structured activity of crafting an authorised rule or programme, applied to a target population of intentional agents within an institution, with the intent of changing a social outcome. Its canonical skeleton is six-part: (1) problem framing — diagnostic of the social pathology to be addressed; (2) goal specification — what outcome counts as success; (3) target population — whose behaviour or which entities' states the policy will affect; (4) instrument choice — the lever (tax, subsidy, mandate, prohibition, information disclosure, nudge, public provision) by which the rule operates; (5) implementation pathway — the administrative and political route by which the instrument becomes enforced behaviour, including street-level discretion and capacity; (6) evaluation and feedback loop — the measurement and revision regime that returns information about effects and drives adjustment. The activity is reflective: a policy designer must anticipate adaptive responses such as gaming and avoidance, second-order distributional effects, and the political economy of authorising and sustaining the rule — the distinction between a well-designed rule and a rule that is actually enforced being what Pressman and Wildavsky (1973) called the implementation gap. Schneider and Ingram's work on target population construction, Howlett on instrument choice, and Sabatier's advocacy-coalition framework each occupy one of the six components as their central analytical object.

Structural Signature

Sig role-phrases:

  • the problem frame — the diagnostic of the social pathology the rule is meant to address
  • the goal specification — the outcome that counts as success
  • the target population — the intentional agents whose behaviour or states the rule will affect, and how they are constructed (deserving/deviant/dependent)
  • the policy instrument — the lever (tax, subsidy, mandate, prohibition, disclosure, nudge, public provision) by which the rule operates
  • the implementation pathway — the administrative and political route by which the instrument becomes enforced behaviour, including street-level discretion and capacity
  • the evaluation loop — the measurement-and-revision regime that returns effects and drives adjustment of the earlier slots
  • the authorising coalition — the political economy that must enact the rule and keep authorising it
  • the reflective anticipation — designed under the assumption that intentional agents will game, avoid, and produce second-order effects, so the implementation gap (rule-as-written vs. rule-as-enforced) is part of the design, not an afterthought

What It Is Not

  • Not mechanism design. Mechanism design is the substrate-independent rule-engineering prime — design rules so self-interested agents produce desired aggregate behaviour — and it spans auctions, matching markets, and biological signalling. Policy design is its institutionalised, publicly-authorised specialisation for rules over intentional agents inside an institution; it composes mechanism design but adds the authorising coalition, the implementation pathway, and the political economy that mechanism design abstracts away.
  • Not the statute as written. Folk reasoning stops at authorisation, treating a passed rule as an accomplished outcome. The discipline insists on the implementation gap: street-level discretion, administrative capacity, and the political economy of sustaining the rule decide whether the text becomes enforced behaviour. A paper-perfect statute that produces no behaviour change has failed at implementation, not in the text.
  • Not a single undifferentiated "policy advice." A policy is a six-part artifact (frame, goal, target, instrument, implementation, evaluation), and the verdict "the policy failed" points nowhere until the break is localised to a component. Collapsing the six into one voice fuses questions that belong to different expertise — economic, legal, administrative, political — and forfeits the diagnosis.
  • Not regulation. Regulation (rule-and-enforcement) is one kind of policy instrument among many — tax, subsidy, mandate, prohibition, disclosure, nudge, public provision. Equating policy design with regulation mistakes a single entry in the instrument slot for the whole design activity.
  • Not the reinforcement-learning or ecological sense of "policy." An RL action-selection rule or an ecosystem's self-regulation borrows the word but drops the defining commitments: no authorising coalition, no intentional target agents who game and avoid, no implementation gap between a written rule and street-level enforcement. The six-part skeleton has nothing to attach to there.
  • Not the decision itself. Policy design is the activity of shaping the option that will be authorised, not the act of choosing it. The commitment is a separate moment; design is the engineering of the artifact that the decision then enacts.

Scope of Application

Policy design lives wherever its defining triad holds — an institution issuing an authorised rule to intentional agents who can adapt — so its habitats are the institutional rule-making settings that preserve that substrate, not a single field; beyond them the word travels but the commitments do not (an ecosystem's self-regulation or a reinforcement-learning "policy" has no authorising coalition, no gaming agents, no implementation gap, and the genuinely portable structure belongs to the component primes it composes — mechanism_design, experimental_design, stakeholder_analysis, institution, institutional_lag).

  • Public policy and government — the home turf: tax codes, welfare programmes, environmental regulation, public-health interventions, and immigration rules, the terrain the six-part skeleton was built for.
  • Corporate policy — HR rules, compliance regimes, code-of-conduct design, and internal incentive structures, where the same factoring and implementation gap apply to firm-internal agents.
  • Educational institutions — school discipline policy, grading policy, and accommodations policy, rules over students and staff who respond strategically.
  • NGO and international development — aid conditionality and programme design at development agencies and foundations, where the authorising coalition and enforcement route are explicit design objects.
  • Standards bodies and self-regulation — industry self-regulatory codes and voluntary certification regimes, authorised rules over member agents who can game or avoid them.

Clarity

Treating policy as a designed artifact with six separable components makes failure diagnosable rather than mysterious. Without the factoring, a policy that does not work invites a single undifferentiated verdict — "the policy failed" — that points nowhere, because the rule, the problem it was meant to address, the population it acts on, the lever it pulls, the administrative route that carries it, and the loop that measures it are all collapsed into one undifferentiated thing. Naming the six components lets the analyst ask which component broke: a sound instrument can fail because the problem was misframed; a precise goal can fail because the implementation pathway was never specified; a well-implemented rule can fail because the target population gamed it. The explicit factoring turns "did it work?" into the sharper "where, in the chain from frame to feedback, did it break?" — and locates the fix at that component instead of scrapping the whole.

The distinction it sharpens most is between the rule as written and the rule as enforced — the implementation gap that Pressman and Wildavsky named. Folk reasoning about policy tends to stop at the statute, treating authorisation as accomplishment; the discipline insists that street-level discretion, administrative capacity, and the political economy of sustaining the rule are part of the design problem, not afterthoughts to it. That lets a designer ask the questions a paper-perfect rule conceals: how will intentional agents adapt to it, who must authorise and keep authorising it, and what does success actually look like when measured. By separating the substantive question (what causes the pathology?) from the instrument question (which lever bites?) from the implementation question (who enforces, with what discretion?), the frame also assigns each to the right expertise at the right moment, so economic, legal, administrative, and political judgment enter where each belongs rather than fusing into a single undifferentiated "policy advice."

Manages Complexity

A policy is the meeting point of four otherwise separate bodies of reasoning — substantive theory (what causes the social pathology?), instrument theory (which lever bites, at which margin?), implementation theory (who enforces, with what discretion and capacity?), and political theory (who authorises, who resists, who must keep authorising?) — and a working rule must get all four right at once. Confronted whole, that is an intractable tangle: a policy that fails returns a single undifferentiated verdict, "it didn't work," that points nowhere because the frame, the goal, the population, the lever, the administrative route, and the measurement loop are fused into one object. The six-part skeleton tames the tangle by factoring it into separately-designable, separately-diagnosable components, so the analyst no longer reasons about "the policy" as a monolith but tracks six things and reads the qualitative outcome off them: a problem frame, a goal specification, a target population, an instrument, an implementation pathway, and an evaluation loop. That factoring is what converts the dead-end "did it work?" into the decidable "where, in the chain from frame to feedback, did it break?" — and the answer localises to a single component with its own branch of remedy, because the failure modes are component-specific rather than diffuse: a sound instrument can fail because the problem was misframed; a precise goal can fail because the implementation pathway was never specified; a well-implemented rule can fail because the target population gamed it. The most load-bearing branch the skeleton makes legible is the implementation gap — the rule as written versus the rule as enforced — which folk reasoning collapses by treating authorisation as accomplishment; isolating implementation as its own component forces the questions a paper-perfect statute conceals (how will intentional agents adapt, what discretion does street-level enforcement carry, who must keep authorising it). And because each component answers to a different expertise — economic, legal, administrative, political — the factoring also routes each question to the right judgment at the right moment instead of fusing them into one undifferentiated "policy advice." So rather than holding an entire ill-defined policy in mind and pronouncing on it globally, the designer tracks six components, the adaptive responses each invites, and the authorising coalition that sustains the whole, and reads off where a failure sits and which expertise must fix it — the move from a monolithic, undiagnosable "policy success" problem to a six-slot decomposition with component-localised failure modes and a sharp written-rule-versus-enforced-rule branch.

Abstract Reasoning

Policy design licenses a family of reasoning moves that all exploit the six-part factoring and the reflective anticipation of intentional agents. Diagnostic: given a policy that did not work, the analyst refuses the undifferentiated verdict "it failed" and instead localizes the break to a component — running the chain from frame to feedback and asking which slot is responsible: a sound instrument that misfired because the problem frame misdiagnosed the pathology, a precise goal undone because the implementation pathway was never specified, a well-implemented rule defeated because the target population gamed it. The signature backward inference is the implementation-gap diagnosis: a statute that is exemplary on paper yet produces no behaviour change is read not as a bad rule but as a failure of enforcement — street-level discretion, administrative capacity, or a lapsed authorising coalition — so the analyst infers from "authorized but inert" to "the gap is in implementation, not in the text." Interventionist: instrument-choice reasoning predicts which lever bites at which margin — when a tax outperforms a mandate, when disclosure outperforms prohibition — and pairs each instrument with its predicted adaptive response, so the designer anticipates gaming, avoidance, and second-order distributional effects before enactment and builds the evaluation loop to detect them; choosing an instrument is thus simultaneously a prediction about how intentional agents will reshape their behaviour around it. Target-population construction is the move in the other direction: the designer reasons forward from how the population is framed (deserving versus undeserving, deviant versus dependent) to the politics the policy will provoke and the durability of its authorising coalition. Boundary-drawing: the frame separates questions that folk reasoning fuses and assigns each to its own expertise — the substantive question (what causes the pathology?) to economic or scientific judgment, the instrument question (which lever?) to mechanism reasoning, the implementation question (who enforces, with what discretion?) to administrative judgment, the authorisation question to political judgment — so a misrouted question (treating a political-economy failure as a technical instrument flaw) is itself a diagnosable error. It also bounds where the concept applies: a rule applied to intentional agents inside an institution is in scope, whereas borrowing the vocabulary for an ecosystem or an algorithm is flagged as the structural commitments not travelling. Order-of-events reasoning fixes the design sequence — frame, then goal, then target, then instrument, then implementation, then evaluation-and-revision — so that a component specified out of order (an instrument chosen before the goal it must serve, an implementation route assumed before the target population is identified) is predicted to produce a characteristic downstream failure, and the feedback loop closes the order by returning effects that drive revision of the earlier slots.

Knowledge Transfer

Policy design's home substrate is broader than a single field — it is any authorised rule applied to intentional agents inside an institution — and within that substrate the six-part skeleton transfers as mechanism, not analogy, because the thing that makes it work (a population of agents who will adapt, an authorising coalition that must sustain the rule, an enforcement route that can diverge from the text) is preserved across institutional settings. So the vocabulary and the diagnostics carry with little loss from public policy (tax codes, environmental regulation, public-health rules) to corporate policy (compliance regimes, codes of conduct, internal incentive structures) to educational institutions (discipline and grading policy) to NGO and development programmes (aid conditionality) to standards-body self-regulation. In each, the same factoring applies: design and diagnose the six components separately; localize a failure to a slot (misframed problem, unspecified implementation pathway, gamed target population) rather than pronouncing "the policy failed"; isolate the implementation gap (rule-as-written versus rule-as-enforced) as its own component so authorisation is not mistaken for accomplishment; anticipate the adaptive response each instrument invites; and route each component's question to the right expertise. The terms — policy instrument, target-population construction, implementation regime, evaluation loop, authorising coalition — keep their referents wherever there is an institution issuing rules to agents who can respond.

Beyond that institutional substrate the transfer is metaphor, and the boundary is exactly where the defining commitments stop holding. "Ecological policy" for how an ecosystem regulates itself, or "policy" in the reinforcement-learning sense of an agent's action-selection rule, borrows the word while dropping what makes policy design a discipline: there is no authorising coalition, no intentional target agents who game and avoid, no implementation gap between a written rule and street-level enforcement. The six-part skeleton has nothing to attach to there — an ecosystem has no problem frame it endorses, an RL policy has no political economy — so importing "policy design" renames the parts and keeps none of the machinery, and should be marked as analogy.

What gives this entry its distinctive shape is that it is itself a domain-specific composition of substrate-independent primes, so the honest cross-domain story is that the portable content lives in those component primes, not in "policy design" as a name. The rule-engineering core — designing rules so that self-interested agents acting under them produce desired aggregate behaviour — is mechanism_design, which travels far wider (auctions, matching markets, distributed-system protocols, biological signalling); the evaluation discipline is experimental_design; the political-economy input is stakeholder_analysis; the authorising frame is institution; and the implementation gap is the rule-versus-enforced-behaviour pattern captured by institutional_lag. Each of those genuinely recurs across substrates as a co-instance, and when a lesson from policy design is wanted outside institutional rule-making, it is one of these parents that should carry it, not the composite. Policy design is the institutionalised, publicly-authorised assembly of those primes for the specific terrain of rules over intentional agents — a high-value composition whose value is precisely that it binds the components together for that terrain, and whose binding does not survive transplant to substrates lacking the institution/intentional-agent/authorised-rule triad. The clean summary: across institutional rule-making (public, corporate, educational, NGO, standards) it transfers as mechanism because the substrate is preserved; beyond it the word travels but the commitments do not; and the genuinely portable structure belongs to the component primes — mechanism design, experimental design, stakeholder analysis, institution, institutional lag — that policy design composes. See Structural Core vs. Domain Accent.

Examples

Canonical

The textbook case is the one Pressman and Wildavsky dissected in Implementation (1973): the U.S. Economic Development Administration's 1966 program to create jobs for unemployed minority residents of Oakland, California. The problem was well framed, the goal (jobs) was clear, generous funding was authorized, and there was broad agreement among the agencies involved. Yet years later almost none of the promised jobs had materialized. The authors traced the failure not to the rule or its funding but to the implementation pathway: the program required a long chain of separate approvals, clearances, and agreements across many actors, and even when each step had a high probability of success, the "complexity of joint action" multiplied the delays until the initiative stalled. Authorization had been mistaken for accomplishment.

Mapped back: Minority unemployment in Oakland is the problem frame and jobs the goal specification; the funded construction projects are the policy instrument. The failure sat entirely in the implementation pathway — the multi-clearance route from authorized rule to enforced behavior — which is the implementation gap the framework isolates as its own diagnosable component rather than blaming the statute.

Applied / In Practice

The U.S. Acid Rain Program under the 1990 Clean Air Act Amendments shows the six components engineered deliberately. The problem was framed as sulfur-dioxide emissions from power plants causing acid rain; the goal was a fixed national cap on SO2; the target population was electric utilities. Crucially, the instrument chosen was not a uniform command-and-control mandate but a market: a fixed number of tradable emission allowances that plants could buy, sell, or bank, letting reductions happen where they were cheapest. The implementation pathway paired this with continuous emissions monitoring and automatic penalties, and a built-in evaluation loop tracked emissions against the cap. Emissions fell substantially and at far lower cost than command-and-control projections, and the program became the reference design for later cap-and-trade schemes.

Mapped back: SO2 acid rain is the problem frame, the emissions cap the goal specification, and utilities the target population. Choosing tradable allowances over a mandate is a deliberate policy instrument decision; continuous monitoring with penalties is the implementation pathway; and cap-tracking is the evaluation loop — each component designed separately and to its own expertise.

Structural Tensions

T1: Separable slots versus coupled components (localization against interdependence). The six-part factoring's whole value is that failure localizes to a slot — misframed problem, unspecified pathway, gamed target — rather than yielding the useless verdict "the policy failed." But the components are not actually independent: the right instrument depends on how the target population is constructed, the feasible implementation route depends on the instrument, and the authorising coalition spans all six at once. So a failure that presents in one slot may originate in an interaction between slots, and the clean localization can misattribute a coupling failure to a single component, prompting a fix to the wrong part. The factoring that makes diagnosis tractable also tempts the analyst to treat as separable what is genuinely entangled. Diagnostic: Does the failure sit in one component, or in the interaction between components (or in the coalition spanning all of them) that the six-slot decomposition tends to hide?

T2: Designed pathway versus street-level discretion (specifying enforcement against the value of judgment). Isolating implementation as its own component forces the questions a paper-perfect statute conceals — capacity, enforcement route, the gap between rule-as-written and rule-as-enforced. The natural remedy is to specify the pathway tightly, minimizing the discretion that opens the gap. But street-level discretion is not only a leak; it is also the adaptive flexibility that lets front-line agents handle cases the rule's authors never anticipated. Design the pathway too tightly and you trade away the local judgment that makes rules workable in the field; leave it loose and the implementation gap widens. The component that names the gap does not resolve the standing conflict between fidelity-to-the-text and responsiveness-on-the-ground. Diagnostic: Is the implementation design closing a gap that genuinely defeats the goal, or suppressing discretion the front line needs to apply the rule to unforeseen cases?

T3: Reflective anticipation versus the adaptation arms race (designing against gaming can provoke it). The discipline's signature strength is treating the target as intentional agents who will game, avoid, and produce second-order effects — anticipating adaptation before enactment. But this cuts two ways. Designing hard against expected gaming tends toward rigid, distrustful, loophole-closing rules that raise compliance cost for the honest and can themselves provoke the avoidance they feared; and no anticipation is complete, so every rule invites a fresh adaptive response the designer did not foresee, launching an arms race the evaluation loop can only chase. The same reflective sophistication that makes policy design more than naive rule-writing can overfit the rule to imagined gaming and calcify it. Diagnostic: Is anticipating the adaptive response producing a proportionate safeguard, or an over-rigid rule whose complexity itself invites gaming and cannot keep pace with agents' next move?

T4: Rational design sequence versus political reality (frame-first against how policy is actually made). The order-of-events logic fixes a sequence — frame, goal, target, instrument, implementation, evaluation — and predicts characteristic failures when a component is specified out of order (an instrument chosen before the goal it serves). This is sound as design discipline, but real policy frequently inverts it: an instrument is chosen first for ideological or coalition reasons, and the problem frame and goal are retrofitted to justify it. The construct can diagnose this as an "out-of-order" error, yet the inversion is not a mistake to be corrected so much as the normal condition of authorised rule-making under political constraint. The rational sequence is a normative ideal that the authorising politics the framework also insists on will routinely override. Diagnostic: Is the design sequence being followed because it is right, or is an instrument already fixed by the authorising coalition with the frame and goal being reverse-engineered to fit it?

T5: Autonomy versus reduction (an institutional composition of portable primes). Policy design is unusually explicit about its own status: it is a domain-specific composition of substrate-independent primes — mechanism_design (the rule-engineering core), experimental_design (the evaluation loop), stakeholder_analysis (the political economy), institution (the authorising frame), and institutional_lag (the implementation gap). Each parent recurs across substrates and is what should carry any lesson outside institutional rule-making; borrowing "policy design" for an ecosystem or a reinforcement-learning policy renames the parts and keeps none of the machinery, because there is no authorising coalition or gaming agent for the composite to bind. The construct's value is precisely the binding — the institutionalized assembly of those primes for rules over intentional agents — which does not survive transplant. The tension is between a genuinely useful composite discipline and the recognition that all its portable structure lives in the components. Diagnostic: Resolve toward the component primes (mechanism_design, experimental_design, stakeholder_analysis, institution, institutional_lag) when carrying a lesson beyond institutional rule-making; toward "policy design" only where the institution / intentional-agent / authorised-rule triad actually holds.

Structural–Framed Character

Policy design sits on the framed side of the spectrum — best read as framed-leaning: a structured human activity, constituted by institutions issuing authorised rules, whose portable content lives entirely in the substrate-independent primes it composes. Four criteria point framed. Its evaluative weight is real if moderate: it is a design discipline oriented toward a rule that succeeds, and it renders diagnoses of failure — though its distinctive move is analytic (factoring, localizing a break to a slot) more than moralizing. It is human-practice-bound in the strongest sense: the concept dissolves the instant its defining triad is removed — no authorising coalition, no intentional target agents who game and avoid, no implementation gap between written and enforced rule — so it does not run observer-free but is constituted by institutional rule-making (an ecosystem's self-regulation or an RL "policy" borrows the word and keeps none of the machinery). Its institutional origin is pronounced: the six-part skeleton, target-population construction, instrument choice, the implementation-gap construct, and the advocacy-coalition framing are analytic furniture drawn inside political science (Pressman-Wildavsky, Schneider-Ingram, Howlett, Sabatier). And vocab_travels fails: policy instrument, authorising coalition, implementation regime, evaluation loop keep their referents only where an institution issues rules to agents who can respond. The one structural-leaning mark is import_vs_recognize: the primes it composes recur across substrates as genuine co-instances, recognized rather than borrowed — but that recognition belongs to the components, not to "policy design," which imports beyond institutional rule-making only by analogy.

Uniquely among these entries, the portable structural skeleton is not one prime but a genuine composition the entry is explicit about — and here more than one is demonstrably needed: the rule-engineering core is mechanism_design (design rules so self-interested agents produce desired aggregate behaviour), the evaluation discipline is experimental_design, the political-economy input is stakeholder_analysis, the authorising frame is institution, and the implementation gap is institutional_lag. Each parent recurs across substrates and is what should carry any lesson beyond institutional rule-making; "policy design" is the institutionalised, publicly-authorised assembly of those primes for the specific terrain of rules over intentional agents, and that binding is exactly what does not survive transplant. So the cross-domain reach belongs to the component primes, while the composite name and its institution/agent/authorised-rule triad stay home. Its character: an institutionally-constituted, practice-bound policy discipline whose every portable element lives in the substrate-independent primes it composes (mechanism design, experimental design, stakeholder analysis, institution, institutional lag), its own value being the domain-specific binding that does not travel — framed-leaning, not a prime.

Structural Core vs. Domain Accent

This section decides why policy design is a domain-specific abstraction and not a prime — and it is an unusual case, because policy design is not a specialization of one prime but a composition of several, so what could lift is not a single skeleton but a set of already-portable components bound together for one terrain.

What is skeletal (could lift toward cross-domain primes). Strip the institution and no single thin structure remains; instead, several genuinely portable primes come apart, each of which the entry names and each of which already travels on its own. The rule-engineering core — design a rule so that self-interested agents acting under it produce desired aggregate behaviour — is mechanism_design, which spans auctions, matching markets, distributed-system protocols, and biological signalling. The measurement-and-revision regime is experimental_design. The who-must-authorize-and-resist input is stakeholder_analysis. The authorising frame is institution. And the rule-as-written-versus-rule-as-enforced gap is the pattern captured by institutional_lag. Because the skeleton here is genuinely doubled — indeed quintupled — each of these is named on its own merit: each recurs across substrates as a real co-instance, and any lesson wanted outside institutional rule-making is one of these parents doing the work. That distributed portable core is what policy design composes, not what makes it policy design.

What is domain-bound. What is specific to policy design is precisely the binding — the institutionalised, publicly-authorised assembly of those primes into a six-slot skeleton (frame, goal, target, instrument, implementation, evaluation) for one terrain, plus the terrain's defining triad: an institution issuing an authorised rule to intentional agents who game, avoid, and produce second-order effects. The worked vocabulary is likewise home-bound — policy instrument, target-population construction (deserving/deviant/dependent), the authorising coalition, street-level discretion, the implementation regime — as are the empirical cases (the EDA Oakland program, the Acid Rain cap-and-trade design) and the field's analytic lineage (Pressman-Wildavsky, Schneider-Ingram, Howlett, Sabatier). The decisive test: remove the authorising coalition, the intentional gaming agents, and the enforcement gap, and the six-part skeleton has nothing to attach to — an ecosystem endorses no problem frame, a reinforcement-learning "policy" has no political economy — so what remains is the component primes standing alone, not policy design.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. Policy design's transfer is bimodal. Within institutional rule-making it travels as full mechanism — the six-part factoring, the localize-the-break diagnosis, the implementation-gap isolation, the anticipation of adaptive response, and the route-each-question-to-its-expertise discipline carry from public policy to corporate compliance to school discipline to aid conditionality to standards self-regulation, because each preserves the institution/agent/authorised-rule triad: genuine recognition of one activity. Beyond that substrate the word travels but the commitments do not: "ecological policy" or an RL "policy" borrows the name while dropping the authorising coalition, the gaming agents, and the implementation gap, so it renames the parts and keeps none of the machinery — analogy, not recurrence. And when the bare structural lesson is needed cross-domain, it is already carried, in more general form, by the component primes themselves — mechanism_design, experimental_design, stakeholder_analysis, institution, institutional_lag — each of which recurs on its own substrate. The cross-domain reach belongs to those parents; "policy design," as named, is the domain-specific binding for rules over intentional agents, and that binding is exactly what does not survive transplant.

Relationships to Other Abstractions

Local relationship map for Policy DesignParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Policy DesignDOMAINPrime abstraction: Experimental Design — is part of, typicalExperimentalDesignPRIMEPrime abstraction: Mechanism Design — is part ofMechanism DesignPRIMEPrime abstraction: Stakeholder Analysis — is part ofStakeholderAnalysisPRIMEPrime abstraction: Institution — presupposesInstitutionPRIME

Current abstraction Policy Design Domain-specific

Parents (4) — more general patterns this builds on

  • Policy Design is part of, typical Experimental Design Prime

    Policy evaluation typically contains an experimental-design discipline for attributing observed effects to the intervention.

  • Policy Design presupposes Institution Prime

    An authorized policy exists only inside an institution whose rules, roles, and enforcement expectations make it binding.

  • Policy Design is part of Mechanism Design Prime

    Rule engineering for adaptive, self-interested agents is the operative instrument-design constituent of policy design.

  • Policy Design is part of Stakeholder Analysis Prime

    Identifying affected, authorizing, implementing, and resisting parties is a constituent of the target-population and political-economy slots.

Hierarchy paths (7) — routes to 6 parentless roots

Not to Be Confused With

  • Mechanism design. The substrate-independent rule-engineering prime — design rules so self-interested agents produce a desired aggregate outcome — spanning auctions, matching markets, and biological signalling. Policy design is its institutionalised, publicly-authorised specialisation: it composes mechanism design but adds the authorising coalition, the implementation pathway, and the political economy that mechanism design abstracts away. Part-versus-whole (one component prime of the composite). Treated more fully in the Knowledge Transfer and Structural Core vs. Domain Accent sections. Tell: is the object a bare incentive-compatible rule for self-interested agents (mechanism design), or that rule wrapped in an authorising coalition and an enforcement route inside an institution (policy design)?

  • Regulation. Rule-and-enforcement — one entry in the instrument slot among many (tax, subsidy, mandate, prohibition, disclosure, nudge, public provision). Equating policy design with regulation mistakes a single instrument for the whole six-part design activity; a policy can pursue its goal through a subsidy or a nudge with no command-and-control rule at all. Part-versus-whole. Tell: is the referent a specific command-and-enforce lever (regulation), or the full frame-goal-target-instrument-implementation-evaluation artifact that might choose any lever (policy design)?

  • Policy analysis. The adjacent discipline of evaluating and comparing policy options — appraising costs, benefits, and trade-offs to inform a choice. Policy design is the generative activity of crafting the artifact that will be authorised, not the ex ante appraisal or ex post assessment of it. Analysis judges options; design engineers them. Tell: is the work weighing and scoring alternatives for a decision-maker (policy analysis), or constructing the six-component rule itself (policy design)?

  • Policy implementation. The administrative and political work of turning an authorised rule into enforced behaviour — street-level discretion, capacity, the multi-clearance route. This is one component (the implementation pathway) of policy design, and the gap between rule-as-written and rule-as-enforced is exactly what the design activity must anticipate. Part-versus-whole: implementation is the fifth slot, not the whole. Tell: is the referent the enforcement-and-delivery phase of an already-authorised rule (implementation), or the upstream engineering of all six components including that phase (policy design)?

  • The decision itself. The act of choosing and committing to the option that will be authorised. Policy design is the activity of shaping the artifact the decision then enacts; the commitment is a separate moment. Design produces the option; the decision selects it. Tell: is the referent the moment of authorising commitment (the decision), or the engineering of the artifact that commitment ratifies (policy design)?

  • Reinforcement-learning / ecological "policy" (the homonym). An RL agent's action-selection rule, or an ecosystem's self-regulation — both borrow the word "policy" while dropping every defining commitment: no authorising coalition, no intentional target agents who game and avoid, no implementation gap between written and enforced rule. The six-part skeleton has nothing to attach to. Pure homonym/analogy. Tell: is there an institution issuing an authorised rule to strategic agents who can respond (policy design), or a decision rule sharing only the word (RL / ecological policy)?

Neighborhood in Abstraction Space

Policy Design sits in a sparse region of the domain-specific corpus (65th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Public Choice & Policy Failure (5 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12