Policy Brief: You Cannot Procure Collective Agency — Protecting the Serendipity-Space and Housing What It Produces

Thought-line: liberty — education — reflection — strike — census-dem Classification: Civic infrastructure + democratic institutional design | Medium-horizon (2–3 political cycles to mature), high path-dependency, self-reinforcing once seeded Decision horizon: The substrate (spaces, rights, slack) can be destroyed in one budget cycle and takes years to regrow; the institution that houses produced agency is cheap to authorize and expensive to authorize credibly. The window is now because the default procurement — “engagement software” — is being signed this cycle in most jurisdictions and it forecloses the alternative.

Problem Statement

A polity wants collective-action capacity: the ability of its residents to notice a shared problem, form a public around it, and act — on housing, heat, a factory closure, a flood. The governing instinct is to deliver this capacity as a service: a consultation portal, a town hall, an engagement KPI, a “have your say” campaign.

That instinct rests on a false premise. Political agency is not a stock the state can dispense; it is produced through action, in real time, and cannot exist prior to it. The improvising collective generates its own reserves — trust, tactics, leadership, obligation — out of shared risk while acting, not before. You cannot hand people a public; they become one by doing something together that none of them could have scripted alone.

Two structural facts follow, and together they define the problem:

The generative mechanism is the catalytic surface. New publics form where overlapping grammars meet — where a tenants’ way of talking collides with a shopkeepers’ with a nurses’ with a congregation’s — and the collision produces encounters no single grammar designed. Agency is a byproduct of unplanned contact across difference. Call the sites of that contact serendipity-spaces.
The dominant governance logic destroys exactly this. The administrative reflex — optimize, script, de-risk, surveil, program every hour of public space toward a measurable outcome — is, in its structure, counterinsurgency: the systematic elimination of serendipity-spaces. It need not be military. Hostile architecture, over-programmed plazas, gig-fractured time, consultation-by-app, and “smart” engagement dashboards all do the same work: they replace unplanned cross-grammar encounter with routed, monitored, single-grammar transactions. They manufacture the appearance of participation while removing its generative substrate.

The policy problem, stated sharply: the interventions that are easiest to fund, measure, and control (engagement platforms, scripted consultations) are precisely the ones that cannot produce agency and often suppress it — because measurability and control are counterinsurgency by another name. The interventions that actually generate collective capacity are low-design and low-measurability by nature (protect a surface; don’t script the outcome), which makes them nearly impossible to defend to a budget office that asks “what’s the KPI?” A decision-maker optimizing on auditability will reliably buy the thing that fails.

One boundary condition governs all of it — the reproductive boundary. The improvising collective can only generate reserves from shared risk if its members have the slack to show up: time, care coverage, the security to withdraw labor without ruin. Agency runs on reproductive labor — the unpaid care and recovery that makes a person available for public life. A population with zero slack (over-worked, precarious, care-saturated) cannot improvise a public no matter how many serendipity-spaces you protect. This is why the thought-line ends at strike and census-democracy: the strike is the paradigm case of agency-produced-through-action, and it is only possible where associational liberty and material slack both exist; census-democracy is the institution that gives the produced agency somewhere to land with real power instead of dissipating.

Who decides. The leadership of a mid-sized city or region with a legitimacy problem — flat turnout, polarization, low institutional trust, and a demonstrated inability to mobilize residents around local shocks. Concretely: a mayor and council (or a regional government’s democratic-renewal / civic-infrastructure office) with a mandate to “rebuild civic capacity,” an annual discretionary budget in the low millions, a two-to-three-cycle horizon, an auditor who wants measurable outcomes, and a chamber where ceding real agenda-setting power to residents is politically fraught and protecting strike/associational rights is charged.

Why now. The counterinsurgency default is not a future risk — it is the procurement on this cycle’s agenda: the engagement-platform contract, the plaza redesign that removes seating, the “activation” program that books every public hour. Each is measurable, vendor-supplied, and controllable, and each one quietly narrows the surface. Once bought, they become the baseline against which the low-design alternative looks like “spending money on nothing.”

Background

Agency is produced, not stored — so it cannot be delivered

The classic error is to treat a “public” as a pre-existing group with latent preferences that better plumbing (a survey, an app) will surface. But publics are called into being by problems and consolidate through the act of responding — the risk taken together is what forges the trust, roles, and obligations that did not exist an hour earlier. This is why disaster-response literature repeatedly finds spontaneous, self-organizing “emergent groups” outperforming pre-planned command structures: capacity was generated in the acting, not drawn from a reserve. The corollary is unforgiving for policy: there is no lever that installs agency directly. You can only protect the conditions under which it self-generates — or destroy them.

The catalytic surface: why difference is the input

Homogeneous, single-grammar gatherings reproduce what their members already know; they do not generate. Novel collective capacity comes from overlapping grammars — the friction of communities that organize reality differently being forced into shared space and having to improvise a common idiom. The “third places” tradition (the café, the library, the market, the stoop — Oldenburg) names the venues; Jane Jacobs named the mechanism (the sidewalk’s unplanned, repeated, low-stakes contact building the trust that later carries high-stakes action). The design implication is counter-intuitive: you protect the surface and refuse to script the encounter. Any program that pre-specifies the outcome has already collapsed the surface into a single grammar — its own.

Counterinsurgency = the elimination of serendipity-spaces

The structural signature of counterinsurgency is not violence; it is the drive to make a population legible and unsurprising — to eliminate the unplanned. Ported into ordinary governance it looks entirely benign:

Physical: hostile architecture, removed seating, over-programmed plazas, enclosure of the commons — each converts a serendipity-space into a transit corridor.
Temporal: gig-fractured, always-on schedules that dissolve the shared, uncommitted hours in which publics form.
Informational: consultation-by-app and sentiment dashboards that route every “voice” through a single monitored grammar and return control to the center.

None reads as repression on the org chart. Each is sold as efficiency, safety, or engagement. Their aggregate effect is the quiet removal of the generative surface — which is why the logic is dangerous precisely because it is administratively respectable.

The reproductive boundary sets the ceiling

Serendipity-spaces are necessary but not sufficient. Participation is metabolically expensive; it is paid for out of slack — discretionary time and the care work that restores capacity for public life. Where slack is near zero, protected spaces sit empty and the assembly seats fill only with those who already had reserves (the retired, the salaried, the un-caring-burdened), reproducing the exclusion the whole exercise was meant to cure. Any serious program must budget for the substrate — time, care, and the liberty to withdraw labor — or it will build a stage no one has the reserves to walk onto.

What we don’t yet know for this jurisdiction (measure before committing)

Where are the surviving serendipity-spaces, and which pending decisions (a redesign, a lease non-renewal, a schedule change) are about to remove one? A rapid catalytic-surface audit is the cheapest high-value first move.
Where is slack concentrated and absent across the population — i.e., who is currently priced out of any produced-agency institution before it opens?
Local frame-risk: how radioactive, here, is “citizens’ assembly with real power” versus “advisory panel,” and “protect the right to strike/assemble” versus “civic space”? Import nothing; test locally.

Options

Option A — Census-Democracy: a standing sortition assembly with real power

Establish a permanent citizens’ assembly chosen by lot (census-based random selection, stratified to mirror the population) with binding or hard-referral power over a defined domain — part of the capital budget, a policy area, or an agenda-setting/veto-referral right — not a purely advisory panel. Real models: the Ostbelgien Bürgerrat (the first permanent sortition body, German-speaking Belgium); Ireland’s Citizens’ Assembly; Paris and Brussels standing assemblies; Oregon’s Citizens’ Initiative Review. What it does in the thought-line: it is the institution that gives produced agency somewhere to land with teeth, instead of dissipating after the moment passes.

Implementability: lower — needs a statute/ordinance, elites must actually cede power for it to work, recurring cost per cycle (member stipends, facilitation, secretariat). Proven templates exist, which de-risks the design.
Expected impact: highest — conditional on real power. Advisory-only degrades into consultation theater and can deepen cynicism when ignored.
Order-of-magnitude cost: a ~~40–60 member body meeting several weekends/year with per-diem stipends (~~€100–150/member/day) plus professional facilitation and a small secretariat runs on the order of low-hundreds-of-thousands of € per year (Ostbelgien’s operating budget sits in roughly this band — verify against local wage/venue costs before committing).

Option B — Protect and seed the catalytic commons (the serendipity-space substrate)

Protect, fund, and stop degrading the physical and temporal spaces of unplanned cross-grammar encounter — libraries, squares, markets, community centers, mixed-use commons, “third places” — and protect the associational substrate: assembly, organizing, and the right to strike. Explicitly reverse counterinsurgency moves: reinstate seating, cap plaza over-programming, roll back hostile architecture, resist enclosure and over-surveillance. The design rule is low-design / high-serendipity: protect the surface, never script the outcome. Real anchors: Ostrom’s design principles for durable commons; the “third place” and Jacobs sidewalk-trust literatures; library-as-civic-infrastructure evidence.

Implementability: medium-high — much of it is cheap or free (stop removing seating; don’t renew a hostile redesign; keep the library open longer). Capital pieces cost more; protecting strike/associational rights is politically charged.
Expected impact: high — this is the necessary substrate. Without it, Option A has no informal feeder and no life between assemblies.
Order-of-magnitude cost: highly variable — from ~€0 (defensive/regulatory moves) to low-millions for capital and extended library/community-center hours.

Option C — Reflective civic capacity + participatory practice

Shift civic education from facts-about-institutions to the capacity to deliberate, reflect, and cooperate: deliberation training, participatory budgeting in schools and neighborhoods (Porto Alegre lineage), and organizing/associational literacy. This is the upstream pipeline that supplies both A (people who can sit in the assembly) and B (people who use the surface).

Implementability: high — comparatively cheap, low-controversy, uses existing schools/community channels.
Expected impact: medium but compounding — slow; pays off over cycles, not months. Weak as a standalone; strong as a multiplier on A and B.
Order-of-magnitude cost: low — mostly curriculum, training, and small participatory-budget pots (tens to low-hundreds of thousands of € depending on the participatory-budgeting envelope).

Option D — Managed digital engagement (the tempting default — named to be rejected)

Consultation apps, e-petitions, sentiment dashboards, “smart-city” engagement platforms, scripted town halls. Vendor-supplied, measurable, controllable.

Implementability: highest — off-the-shelf, cheap-ish, generates dashboards an auditor loves.
Expected impact: near-zero, plausibly negative. This is the counterinsurgency option in participation’s clothing: it routes every voice through one monitored grammar, eliminates serendipity, produces compliance-data not agency, and breeds cynicism when the “input” is visibly ignored. It scores best on exactly the axis (measurability/control) that selects against the thing we want.
Why it’s on the list: because it is what most governments actually buy, and naming it is the point — it is the trap the thought-line diagnoses.

Trade-offs

Criterion	A: Census-democracy	B: Catalytic commons	C: Reflective capacity	D: Digital engagement
Implementability (cost, speed, statute, auditability)	Low–Medium	Medium–High	High	Highest
Expected impact (real agency generated)	Highest*	High	Medium (compounding)	Near-zero / negative
*Generates produced* agency (vs. staged)?**	Yes*	Yes	Indirect	No
Auditor-friendly (clean KPIs)?	Partial	Low	Medium	High (deceptively)
Depends on reproductive slack?	Heavily	Moderately	Moderately	Barely (that’s the tell)
Political risk	High (ceding power)	Medium–High (strike/enclosure fights)	Low	Low
Reversibility if wrong	Medium	High	High	High
Failure mode	Advisory-only → theater	Under-funded → empty spaces	Too slow alone	Manufactures compliance, kills serendipity

* Option A’s impact is conditional on real (binding or hard-referral) power. Advisory-only collapses it toward Option D.

Ranking — implementability: D > C > B > A. Ranking — expected impact (real agency): A > B > C ≫ D. Combined (impact, weighted for the actual goal, discounting the auditability that misleads): B and A lead as a pair; C multiplies them; D is a negative-value distractor that will nonetheless win any decision made on auditability alone.

The decisive trade-off: the option that best satisfies the budget office (D) is the one that cannot work and can backfire; the options that work (B, then A) are the hardest to defend on a KPI sheet because their virtue is unscripted, unmeasured encounter. Choosing on measurability is choosing counterinsurgency. The entire brief turns on refusing that.

Second-order effects:

A without real power doesn’t just underperform — it actively teaches residents that participation is decorative, which is worse than not asking. Power is the load-bearing variable.
B protected but slack-starved builds a stage no one can afford to stand on; the seats fill with the already-resourced and reproduce the exclusion. Fund the reproductive boundary or B is cosmetic.
D crowds out. Once “we have an engagement platform” is the answer, the pressure to protect the surface evaporates — the trap’s worst effect is that it feels like the problem is solved.

Recommendation

Do not procure agency. Protect its substrate (B) and build the institution that houses what the substrate produces (A), with reflective capacity (C) running throughout — and explicitly reject Option D as the default even though it wins every auditability contest. Fund the reproductive boundary (slack: time and care) as the enabling condition, or none of it works.

Sequence, because these are complements, not substitutes — they are the thought-line made operational (liberty → education → reflection → strike → census-dem):

Now, near-zero cost — run a catalytic-surface audit and stop the bleeding (B, defensive). Map surviving serendipity-spaces and the pending decisions about to remove one (a redesign, a lease, a schedule, a hostile-architecture install). Freeze the removals this cycle. This is the cheapest, highest-leverage move available and it buys time. It is also pure liberty-substrate: protect the surface before optimizing anything.
This cycle — install census-democracy with real teeth (A), or don’t bother (A). Authorize a permanent sortition assembly with binding or hard-referral power over a bounded domain (start with a slice of the capital budget — the Porto Alegre lesson is that money decisions make participation real). Copy Ostbelgien’s permanent design rather than a one-off. If the chamber will only grant advisory status, defer A and put the money into B — an advisory body is Option D with chairs.
Throughout — fund the reproductive boundary so the seats are reachable. Attach stipends, care coverage/childcare, and scheduling that assumes shift and care workers to every participation venue (assembly, commons, PB process). Protect associational and strike liberty as the material form of “agency produced through action.” Without slack, A and B are stages for the already-comfortable.
Over cycles — seed reflective capacity (C) via deliberation training and neighborhood/school participatory budgeting, as the pipeline feeding A and B.
Governance guardrail — protect the unmeasured; refuse the counterinsurgency reflex. Write into the enabling language that the assembly’s power is binding/referral, not advisory, and that commons funding is not conditioned on scripted-outcome KPIs. Report B and C in inputs protected (spaces kept open, hours funded, hostile removals reversed, slack disbursed), never in engagement-dashboard metrics — because the moment you manage the surface to a number, you have re-scripted it into a single grammar and killed the serendipity. The refusal to over-measure is not laxity; it is the design.

What to do first (90 days):

Catalytic-surface audit + removal freeze (B-defensive) — cheap, fast, protects optionality.
Draft the sortition ordinance with the power granted up front (A) — power is the one thing that cannot be added later without re-earning trust.
Local frame-risk test on “assembly with real power” vs. “advisory,” and on the strike/associational language — test here, import nothing.
Cost the reproductive-slack line (stipends + care + scheduling) and put it in front of the assembly budget, not as an afterthought — the honest arithmetic is that participation without paid slack is participation by the already-resourced.

What remains undetermined. Whether an institution (A) can host genuinely produced agency or whether housing it inside a formal, funded, facilitated body already re-scripts it — domesticates the improvising collective into one more managed grammar. The honest read: the risk is real, and the only hedge is to keep A coupled to an unmanaged B. The assembly must stay porous to the street it draws from; a citizens’ assembly with no living commons behind it is a battery with nothing charging it. Protect the surface first; the institution is only as alive as the serendipity still feeding it.

You cannot hand a population its own agency. It makes itself, in the act, out of shared risk — on a surface of overlapping grammars that no authority designed and every authority is tempted to optimize away. The policy question is not “how do we engage them?” — that question is already counterinsurgency, already the single grammar routing every voice back to the center. The question is: which surfaces will we protect from our own urge to manage, and will we give what forms on them somewhere real to land? Protect the space; fund the slack; grant the power; then get out of the way.