To everyone who has been told that public decisions are too important to experiment with,
I agree with the warning more than you might expect. A community is not a laboratory full of objects. Its residents are people with memories, obligations, vulnerabilities, and the right to refuse a careless test. That is why I have spent my working life designing experiments in governance: not to make people into subjects, but to make the rules that govern them answerable to evidence.
My name is Mara Vale. I work in a fictional coastal municipality called South Reach, where the council once treated every failed policy as a public embarrassment and every successful policy as proof that further questions were unnecessary. We had elections, hearings, committees, and a thick volume of procedures. We also had an impressive ability to repeat a mistake without learning from it.
The first experiment I proposed concerned the public meeting itself. Attendance had fallen for eight years. The council blamed apathy. Residents said meetings were held when they were working, that speakers were interrupted, and that decisions appeared settled before anyone reached the microphone. The chair insisted that every citizen enjoyed the same right to attend.
He was technically correct. The doors were open. The practical conditions of entry were not equal.
I suggested a trial. For three months, one committee would hold alternate meetings in the evening, publish short plain-language summaries before debate, offer remote testimony, and provide a supervised children’s room. Another committee would continue its existing process, while both recorded not only attendance but whose concerns were addressed and whether participants returned.
The chair disliked the proposal. He asked why we needed an experiment when the existing rules were lawful. I replied that legality answered a different question. A procedure can comply with its own requirements and still fail to make participation possible.
We designed the trial carefully. No resident would lose an existing right. Anyone could attend either format. We would disclose the limits of the comparison, because committees dealt with different issues and attendance could rise for reasons unrelated to our changes. We would not pretend a small trial was a scientific proof of universal effectiveness.
At the first evening meeting, a warehouse worker arrived still wearing her uniform. She spoke about a dangerous crossing near the bus depot. The issue had appeared in written complaints before, but no one had connected it to the shift change that sent hundreds of workers across the road after dark. A councillor asked for a site visit. The worker looked startled. She said she had never before been in the room when the council discussed her street.
That was an encouraging moment, not a final result.
The trial also revealed problems. Remote testimony failed for residents with unreliable connections. The children’s room was understaffed one evening. Some older residents preferred the daytime schedule. A meeting that ran late excluded people who depended on the last bus. Our first proposal had not solved participation. It had exposed the many forms participation could take.
We revised the design instead of declaring victory.
Governance experiments are often sold with the language of innovation, as if novelty were a moral virtue. I have become suspicious of that language. A new process can be worse than an old one. A digital portal can remove a queue from a hallway and place a longer queue inside people’s homes. A rotating committee can broaden participation or merely move responsibility so quickly that nobody learns enough to act.
The question is not whether an idea is new. It is whether the people affected can see what is being tested, challenge the assumptions, and help decide what should happen next.
Our second trial concerned maintenance requests. South Reach had a form that asked residents to select a problem category. Street lighting, drainage, footpaths, and waste collection appeared in separate menus. A woman reported that water from a broken drain collected around an unlit crossing. Her complaint was rejected twice because she had selected the wrong category. The third submission was accepted as drainage, then closed when the drain was cleared. The light remained broken.
Each department had fulfilled its assigned task. The crossing remained unsafe.
I asked whether one small neighbourhood could try a shared case number for problems involving multiple departments. The trial required no new surveillance and collected no additional personal information. Residents could still submit separate complaints if they preferred. The difference was that a case would remain open until the linked issues were reviewed together.
Some managers feared that the experiment would damage their performance scores. Their concern was rational. They had been rewarded for closing tickets, and a shared case could remain open longer. We changed the evaluation criteria before the trial began. Otherwise, the experiment would have asked staff to improve outcomes while punishing them for doing so.
This is the lesson I return to most often: you cannot test a better process fairly while leaving the incentives of the old process untouched.
The shared cases revealed that a small number of locations generated repeated complaints across several departments. We had not discovered a secret conspiracy. We had discovered a design problem in the way responsibility was divided. A repaired drain, an adjusted light, and a safer crossing could be coordinated at a cost lower than the repeated visits we had been making.
Not every shared case was easy. Some involved private property, budget limits, or conflicting needs. But residents could now see which question remained unresolved. The council could no longer count a problem as solved simply because it had moved out of one office.
The success attracted attention. A regional agency offered funding if we expanded the model immediately to every district. I declined. People called me cautious, which amused my colleagues. An experimenter is expected to be impatient with caution.
But scaling a pilot too quickly can erase the conditions that made it work. Our trial relied on two experienced coordinators who knew the neighbourhood. We needed to learn whether the method could survive staff turnover, larger caseloads, and different local needs. We also needed to know whether the extra coordination burden was fair to the workers doing it.
An experiment without a plan for failure is not an experiment. It is a publicity campaign with a spreadsheet.
Before each trial, we now write down what would make us stop. We name foreseeable harms, identify who might bear them, and establish a way to report problems without retaliation. We ask whether participation is genuinely voluntary and whether the experiment affects essential services. When a proposal could deprive someone of a basic right or necessary support, we reject the idea of random withholding and choose a safer evaluation method.
Sometimes the ethical design is to compare a new service against the current one while guaranteeing a minimum standard to everyone. Sometimes it is to phase an improvement in as resources become available and learn from the rollout. Sometimes the evidence is already strong enough that another experiment would merely delay action.
Governance does not become ethical simply because it uses the word evidence.
Our third trial failed. We created a rotating residents’ panel to review minor park improvements. The selection process was transparent, and members received modest compensation for their time. We expected fresh perspectives. Instead, the panel spent most meetings deciphering technical plans. Members felt embarrassed asking basic questions, and staff grew frustrated repeating explanations. By the final meeting, several participants had stopped attending.
The first instinct of the council was to blame the residents for not taking the role seriously. The second was to blame staff for insufficient support. Neither explanation captured the design flaw. We had invited people to decide without giving them the knowledge or time needed to exercise that responsibility.
We published the failure. It was one of the most useful documents our office ever produced.
For the next attempt, we provided accessible briefings, paid preparation time, and a clear distinction between questions requiring technical expertise and questions involving public priorities. We also allowed participants to request an independent explanation. The panel did not become infallible. It became capable of making an informed choice.
The difference matters. Participation is not the ceremonial presence of a citizen beside an expert. It is a relationship in which knowledge is shared sufficiently for judgment to become possible.
South Reach has people who distrust every experiment. Some remember reforms that promised empowerment and delivered new paperwork. Others fear that temporary trials will become permanent without public consent. Their scepticism is not an obstacle to be overcome by clever messaging. It is information about the history of trust in our town.
We therefore publish the purpose, duration, budget, decision criteria, and possible consequences of each trial before it begins. We identify who can pause it. We report negative outcomes alongside positive ones. We invite affected residents to contest our interpretation of the results. We do not allow a favourable average to erase severe harm concentrated among a smaller group.
One councillor once asked whether all this disclosure made experimentation too slow. I told him that the purpose of speed was not to arrive at a mistake sooner.
Of course, delay also has a cost. People should not be forced to wait indefinitely while officials request another study. We set deadlines for decisions and explain why more evidence is necessary when we ask for it. Sometimes the honest answer is that we must act under uncertainty and monitor what follows.
I think often of An Open Letter on Governance by Default, which reminds us that doing nothing preserves an arrangement with its own effects. I also revisit A Letter from a Central Planner, because the view from above can make a model appear more complete than the world it describes. Experiments should challenge that illusion, not strengthen it.
People ask whether South Reach has become better governed. I cannot offer a single number. Some services have improved. Some proposals were abandoned. A few conflicts became more visible because people finally believed that speaking might matter. Our meetings are occasionally longer and less comfortable. Our reports contain more uncertainty than they used to.
Those are not all signs of failure.
A healthy governing institution does not need to be perpetually in motion. It needs the ability to distinguish a stable practice worth preserving from a habit protected only by its age. It needs the humility to test its assumptions and the discipline to stop when the test itself becomes harmful.
If you hold authority, begin with a problem people actually experience, not a technology you are eager to display. State what you believe is happening and what evidence would change your mind. Invite the people who bear the costs to help define success. Give staff enough time and resources to participate honestly. Protect the right to object.
If you are governed by such an institution, you should be able to ask who chose the question, who chose the measure, who can stop the trial, and who will live with the outcome. You should receive answers that do not require a professional translator.
One evening the warehouse worker who first spoke at our trial returned to a meeting. The crossing near the depot had been redesigned. She thanked the road crew, then asked why the new pedestrian signal was too short for a colleague who walked with a cane. Nobody accused her of ingratitude. The engineer made a note and agreed to measure the crossing time with the people who used it.
That was the moment I understood what success might look like. Not a finished system. A system that could hear the next question without treating it as a threat.
We will keep making mistakes. We intend to make fewer of them in secret.
With respect for the people whose lives are never merely experimental,
Mara Vale, governance experimenter of South Reach



