Your neighbour knocks: "The tree on your lot is cracking the shared wall." A responsible reply isn't "Thanks for the feedback," but asking: who will inspect, when, what they'll do if risk is high, how to appeal if the fix fails, and what you owe if a fix is late.
Listening without acting is theatre. Acting without limits is arbitrary. Responsibility bridges the two by turning recognition into engagement with teeth — promises you can verify, contest, and revoke.
Tronto elevates responsibility as the most politically central phase of care. She reframes democratic politics itself: not "who gets what, when, and how" (Lasswell) but "who is responsible for caring for what, when, where, and how." The shift replaces a distributive frame — who wins the allocation contest — with a relational one: who has made a binding commitment, to whom, and how do we know if they have kept it?
Why it matters
Why must responsibility be explicit? Because power defaults to evasion — what Tronto names privileged irresponsibility: the power to define which care obligations apply to you, and the habit of choosing the ones that cost least. She identifies five structural "passes" through which the evasion operates:
- The protection pass. Those who protect (military, security) claim exemption from other care.
- The production pass. Those who earn claim exemption from household and community care.
- The taking-care-of-my-own pass. Those who care intensely for their own children or group claim exemption from caring about others.
- The bootstrap pass. "You should have arranged your own care through the market."
- The charity pass. "Voluntary giving is enough; no collective obligation is needed."
Corporations play the same game at scale — diffusing authority until no one is answerable — which is why Tronto calls the neoliberal state itself an "irresponsibility machine" that cranks out one standard answer: "They're your own. You're on your own." These passes are the irresponsibility machine that Civic AI must short-circuit — and AI extends the machine with passes of its own: complexity (no single person can answer for so intricate a system), distribution (responsibility signed away at each joint between trainer, deployer, and licensee), and speed (the system outran governance, so the moment for accountability passed). Engagement contracts exist to make passes visible and revocable. One pass specific to AI is the community knowledge pass: treating community knowledge — local traditions, language, tacit expertise — as free input, a resource to be extracted rather than a contribution to be compensated. When a Kami — knowledge artefact management intelligence — relies on that knowledge to function, the communities maintaining it are contributors whose labour the contract must recognise and compensate.
Definition
- Answerability is the unit. If no answer is required, nobody is responsible.
- Authority must match duty. No duty without the powers (budget, access, pause) to fulfil it.
- Promises over preferences. Goodwill is fragile; verifiable commitments travel.
- Formalise without legalism. Contracts record obligations in public form, but later packs still test, revise, and repair them.
- Two lanes. Big changes follow the full contract; small, safe-to-fail bets run in a sandbox with tighter bounds and quicker cycles.
- Institutional memory. Making responsibilities fractal — mirrored from team to agency — ensures they survive leadership changes.
Core artefact: the Engagement Contract
Every significant deployment carries a published Engagement Contract — a short, legible spec anyone can audit: closer to a well-designed health and safety notice than a terms-of-service agreement. The contract makes obligations public; it does not replace the continuing judgment required to interpret, revise, and repair them.
Four headings to remember (one page if possible):
- Scope. Purpose and non-purpose; data inputs and outputs; retention; deletion on handover; rights baseline and non-negotiable guardrails.
- Obligations. Severity classes and service levels; the accountable person; what adopt-or-explain requires when Assembly outcomes are not followed.
- Brakes. Pause and rollback triggers; who can invoke them (PO, oversight, quorum of affected people); how long emergency powers last.
- Remedies and record. Correction, rollback, compensation; pre-funded remedies; a tamper-evident change log; conflicts disclosed; contacts kept current.
Oversight with teeth
- Independent board (community + domain + legal). Can pause or veto high-impact changes; must publish reasons and declare conflicts.
- Protected budget & terms. Resists capture.
- Open docket. Anyone can file a challenge; triage is public; decisions are reasoned.
- Clawbacks & penalties. Breached promises trigger automatic remedies (escrow drawdown, withheld payment, probation).
- Public attestation. Weekly contract state and diffs are digitally signed and mirrored to a public registry.
From ideas to practice
- Translate recognition into a spec. Convert attentiveness outputs into an Engagement Contract.
- Assign a Participation Officer (PO). Task the PO with running the promise loop, tracking the ledger, and escalating.
- Wire brakes before launch. Incorporate role-based pause/rollback buttons; test them.
- Pre-fund remedies. Pre-fund escrow for compensation and rollback costs at the highest severity; mutual insurance pools or automatic pause for lower tiers — tier by impact, not organisational form.
- Tie payment to proof. Keep vendor pay linked to promise delivery — SLA adherence and adopt-or-explain rate — not raw engagement.
- Run adopt-or-explain. Integrate Assembly outputs or publish a reasoned deviation + remedy.
- Attest & publish. Use independent audits to compare behaviour to contract; hash the diffs to a public mirror.
- Handover or shutdown. Hand off with full records when scope ends — or trust breaks — or switch off gracefully.
One case: the flood-bot
After the flood, the city's flood-bot must pay people on time and fix mistakes.
- Contract. Contract promises livelihood cases decided in 48h; pause if denials spike >25% in any district; rollback if appeals on a rule exceed 20%.
- Owner. A named PO publishes the obligation ledger and signs weekly attestations.
- Adopt-or-explain. The Assembly endorsed multiple proofs of residence. The team adopts three (utility bill, employer letter, neighbour attestations) and explains excluding bank statements (exclusion risk), offering a kiosk-notarised sworn statement as remedy.
- Pause. Night-shift district shows 31% denials in 24h — Elena's claim among them, rejected for lacking a paper lease. The oversight board hits pause; the older "30-day proof" rule rolls back; emergency disbursements use a reversible default.
- Remedies. Wrongly denied claims get automatic compensation (late fee + apology + fast-track). Escrow funds claims the same day.
- Handover clause. The contract stipulates a strict six-week sunset and handover: records and models transfer to the housing office; the switch-off is logged; the ledger is archived. The terms are now set. Execution comes when the clock runs out.
A second case: the deepfake liability flywheel
The 2024 Taiwan deepfake-scam Assembly — 447 citizens at 44 tables of roughly ten — was itself an Engagement Contract at civic scale.
- Contract. Liability sits where the capability lies: a platform shares the loss when an unsigned advert impersonates someone and a citizen is defrauded. Citizens cannot audit algorithms; platforms can verify signatures — so the contract reversed the direction of liability.
- Brake. A platform that fails to keep a local representative able to handle liability claims has its video traffic slowed by 1% for each day of non-compliance — a graduated brake, not a ban.
- Result. The package earned roughly 85% support among the deliberants and passed Parliament within months. Taiwan's Ministry of Digital Affairs (moda) reported identity-impersonation scam ads down roughly 94% in the categories the law touched — a near-collapse, but a targeted one.
- Flywheel. A Reuters investigation found that Meta simply rerouted the affected adverts to nearby jurisdictions without the same rules — harm displaced, not ended — and Japan is now fast-tracking equivalent legislation. One country's fix lowers the cost of the next.
What could go wrong
- Scope creep. Bot starts "screening fraud" unrelated to relief. Fix: Enforce outer bounds; require fresh authority for scope changes.
- Responsibility ping-pong. Teams blame each other. Fix: Single named owner per promise; PO escalates stalled dependencies.
- Paper commitments. PDFs with no force. Fix: Escrow, clawbacks, attestation and audit triggers wired before launch.
- Unfunded mandates. Duties without budget. Fix: Block go-live unless authority and funding match duty.
- Quiet rollbacks. Rules change without notice. Fix: Mandatory public diffs; unlogged changes are invalid.
- Escrow gates out the grassroots. Only well-funded actors can post bonds, re-centralising AI. Fix: Tier liability by severity; mutual insurance for community deployments; pause triggers replace escrow where financial stakes are low.
Interfaces
- From Attentiveness (Pack 1): who/what/why arrives with rights flags and uncertainties.
- To Competence (Pack 3): responsibility turns needs into specs, SLAs, and brakes — safe-to-fail by default.
- To Responsiveness (Pack 4): remedies, rollbacks, and evals are routine; repair is part of delivery.
- To Solidarity (Pack 5): portability and exit rights (detailed in Pack 5) are referenced in every contract.
- To Symbiosis (Pack 6): bounded scopes, handover, and shutdown are success criteria — a Kami that cannot end gracefully has failed the care test.
A closing image: the signed work order
Picture a work order by the door: what will be fixed, by whom, by when; how to check the work; who to call if it fails. The signature is legible — and so is the penalty for not showing up. In other words, teach our systems to post their work orders, sign them, and honour them — and design the oversight that makes honouring them the path of least resistance, not an act of heroic institutional will.
