civic.ai仁工智慧
華文
Pack 4 · 回應力

Pack 4: Responsiveness — Care-Receiving

Whether affected people can contest outcomes and force repair.

Hand-lettered comic page headed 'RESPONSIVENESS — PROBLEM': a rabbit holding a compass explains that society needs metrics to know where it is going but those metrics are often broken; paired charts contrast a rising 'average income' with the reality that income soars for the upper class while stalling for the lower class, and a table maps clicks to clickbait, comments to rage-bait flame wars, and user approval to sycophantic chatbots.RESPONSIVENESSPROBLEM:Like how navigators need a compass, society needs a way to measure where we're going. So, we use metrics! But alas, metrics are often broken.For example, policymakers may say that "average income” is up......but "average" hides the fact that income is way up for the upper classes, while stalling or falling for the lower classes.Likewise, the algorithms & AIs that run modern media -- and thus run modern culture & politics -- use metrics with famous failures.MetricFailureclicksclickbaitcommentsrage-bait flame warsuser approvalsycophantic chatbotsetcSo how can we measure what we value -- our "compass" --without being mislead?

A clinic posts hours, yet patients show up to a locked door. A responsive clinic apologises, posts why it happened, updates hours, and texts people next time. The fix becomes part of the system.¶ 1

Competent action creates new information. The person who was served, or failed, now knows something about the system that the system does not yet know about itself. Refusing to receive that knowledge is the fastest path to repeated failure.¶ 2

In Caring Democracy, Joan Tronto defines responsiveness as “observing that response, and making judgments about it (for example, whether the care given was sufficient, successful, or complete?)” She adds a point that is easy to overlook: the person cared for need not be the one who responds; others in the care setting can assess. Those best placed to judge whether care worked are usually the people who live closest to its effects.¶ 3

Aimee van Wynsberghe read the same element into design a decade ago: responsiveness is what “encourages an active stance of the care-receiver rather than a more passive, vulnerable one” (2013, p. 418). For Civic AI, the public question is simple: when a system harms someone, who can contest the result, how fast, and what changes because they spoke?¶ 4

Start in your room

Imagine saying to an institution, “I said this was wrong. Who can change it, and when will they answer?” Make it easy to reach a person who can review the case, then follow it through until the repair is made.¶ 5

Definition

Why it matters

Pack 3 checks whether the process ran as promised. Pack 4 checks whether the care actually landed. So responsiveness begins with community judgement.¶ 11

An aggregate satisfaction score, or any single number, cannot replace situated knowledge. Surveys can detect large shifts in sentiment, but they are structurally blind to what a night-shift nurse or a community member notices at the edge of the system. It takes both the numbers and people with authority to act on what they see.¶ 12

Community-authored evaluations, appeals, and repair logs are the public commitment. Some teams later feed those signals into routing or model updates (sometimes called Reinforcement Learning from Community Feedback, RLCF), but that choice comes second. The community writes the yardstick first.¶ 13

Responsiveness keeps the other three phases from becoming control. Care without the care-receiver’s own assessment projects the carer’s preferences onto a passive recipient. Institutional care has caused enormous harm this way: care that was actually control. The residential schools that separated Indigenous children from their families were, by many official accounts, expressions of care. The psychiatric institutions that confined people judged deviant often understood that confinement as therapeutic.¶ 14

A system that is attentive, responsible, and competent, but has no mechanism for the people it serves to say “this is not working for me,” can drift, undetected, from care into domination. The person receiving care, or someone authorised to speak for them where they cannot, must be able to say that sentence, and the sentence must change the system.¶ 15

The brake as a care mechanism

Feedback pipelines can change future training runs, but they cannot interrupt a wrong decision in the moment when interruption is what care requires. That takes a brake: the mechanism by which the care-receiver, or the care-giver on their behalf, can say “this time, the automated model is wrong; stop, let me override, or route this to a human now.”¶ 16

For a nurse, that means overriding the system’s “no alert” verdict, stopping it from closing the case, and calling the doctor without justifying it to a software interface; for a resident, a two-minute action that triggers human review of their case; for a language community, a way to flag a pattern and force a policy review, not merely lodge individual complaints. A brake works when people can reach it without a maze and pulling it actually stops something.¶ 17

An honest design cannot hide the tension: a brake that is easy to reach is also easy to lean on, and a ward where every alert is overridden has not solved alarm fatigue. It has automated it. The answer is a legible brake, not a harder one: log every pull in an attributable running record, the override ledger, and feed over-override patterns back into community-authored evaluations as a signal in their own right.¶ 18

What it looks like in practice

From ideas to practice

  1. Expose the appeal button. Every decision should show a one-click route to challenge it, with the clock visible.¶ 24
  2. Accept harm drafts. Let people describe harms in plain language; turn them into community-reviewed tests with partners.¶ 25
  3. Triage by severity. Highest-severity cases trigger immediate pause or reversible defaults.¶ 26
  4. Fix or explain. Publish the remedy or the reason with next steps, on the clock.¶ 27
  5. Memorialise. Turn incidents into tests and link them from the contract changelog.¶ 28
  6. Check back. Close the loop with the people who appealed and ask whether repair made the process more trustworthy. Our proposed trust-under-loss test is informed by Tom Tyler’s procedural-justice research. It is not a validated AI measure or a promise that people who lose will trust the system.¶ 29

Buildable tools

One case: the flood-bot

This is a fictional composite for illustration. It is not operating evidence.¶ 36

What could go wrong

Interfaces

A closing image: the workshop wall of retired broken parts

Imagine a workshop with a wall full of retired broken parts, each tagged with the story of how it broke, how to avoid future damage, and who fixed it. The wall is a wall of learning, not shame. It is public, and kept by the community that uses the workshop as well as the engineers who built the machines. The shop that hides its breaks will repeat them. The shop that builds the wall becomes, over time, the most reliable shop in the district, because it has made its failures the foundation of its competence.¶ 53

Sources for this pack

Hand-lettered comic page headed 'NOT: metrics that mislead us, BUT: metrics made by us, for us — SOLUTION': three steps — measure what we actually care about, measure it in multiple ways, and re-evaluate metrics often — illustrated with Weval.org, a public collection of AI-chatbot evaluations measuring values like honesty, cultural fluency, conflict resolution and mental-health best practices, with green 'BY US' and 'FOR US' arrows rising above a crowd of animal characters.not: metrics that mislead us but: metrics made by us, for usSOLUTION:1. Measure what we actually care about. 2. Measure it in multiple ways. 3. Re-evaluate metrics often.Consider Weval.org, a collection of evaluations for AI chatbots, designed by & for the public!While industry mostly tracks AI capabilities, Weval measures what people value: - honesty - cultural fluency - conflict resolution - mental health best practices ...and so on!Each value is measured in many different ways, for robustness.And there's a feedback loop: weevaluate the evaluations,& improve them.BY USFOR USThis is how we can design value-based "compasses" to guide our technology!

←Pack 3: Competence
Pack 5: Solidarity→Back to Pack 1: Attentiveness

The Trace

Amended
Stewards
Audrey Tang · Caroline Green
with jdd-kami · Tenzin Yangtso
Amended
· 52 recorded amendments since 1 Oct 2025
Full amendment ledger →
Source
4.md · edit on GitHubhttps://github.com/audreyt/civic.ai/blob/main/4.md
CC0 public domainhttps://civic.ai/4/
Evals
links ✓ passparity ✓ passterms ✓ passcontrast ✓ passsize ✓ passBuilt 2026-10-13T21:41:55.323Z · Full eval receipts (JSON)
Contest
Disagree? File an amendment.
Write to the stewards · [email protected]
Twin
四：回應力——接收關懷
Same record, same receipts.