A handbook for testing new power systems, medicines, vehicles, building materials, and public machines without making ordinary people carry an experiment's risks alone.
A civic experiment leaves the laboratory when ordinary people begin depending upon it.
An experimental power regulator may darken a clinic. A new paving material may become slick in rain. A transit signal may direct two vehicles into the same lane. A medicine may cause a reaction that never appeared in a small trial. A service automaton may follow an instruction correctly and still injure someone its designers failed to consider.
The public did not enter the laboratory. The experiment entered public life.
Responsibility must be assigned before that happens.
Before the trial
Every proposal names the office responsible, the lead examiner, the workers allowed to stop the trial, the people likely to be affected, the expected benefit, the known risks, and the conditions that require a return to the old system.
“The laboratory” is not a responsible name. A building cannot answer questions, repair a street, treat an injury, or pay compensation. Name the people and offices that can.
Test individual parts under controlled conditions before connecting them to a public system. Heat the paving sample. Freeze it. Wet it with clean water, oil, salt, and the waste commonly found on the street where it will be used. Load a bridge joint until it fails. Run a pump with dirty intake water. Interrupt the power and observe how the machine restarts.
Passing the expected test is not enough. Public life supplies conditions the designer did not request.
Power and water trials
Begin with one building, pump, light circuit, or storage system that can be isolated quickly. Keep the previous equipment ready until the new system has survived normal load, peak load, poor weather, maintenance, and failure.
A backup that has never carried the full load is an untested machine, not a safety plan.
Tell residents when interruptions may occur and where emergency water, heat, light, air, or charging can be found. If a public office creates the interruption, the office provides the temporary service. People should not have to purchase protection from a test they did not request.
Medicine and clinic trials
Explain what is established, what remains uncertain, what alternatives exist, and what support will be provided if harm occurs. Payment for participation does not remove the right to leave.
Begin with the smallest group able to answer the question. Do not exclude difficult bodies merely to produce clean results, then sell the treatment as suitable for everyone. Differences in age, species, implants, metabolism, pregnancy, prior injury, and other medicines may change the result.
A patient receiving ordinary care should not discover afterward that a new medicine, instrument, or procedure was being tested upon them. Emergency treatment may require quick decisions. It does not make every experiment part of the emergency.
Transport trials
Test new signals, schedules, lifts, brakes, route controls, and traffic patterns when the smallest number of travelers will be exposed. Place trained workers where they can see failures directly and restore the previous route.
Provide another way through whenever possible. A traveler who cannot use the experimental lift still needs to reach the platform. A district should not be cut in half because a route designer wants uncontaminated results.
Count near misses, delays, frightened animals, mobility failures, confused signs, and workers forced into unsafe improvisation. A trial did not succeed merely because no vehicle completed the collision it almost began.
Materials trials
A material approved for one use is not approved for every use. A wall panel may resist flame and release poisonous smoke. A strong floor may become brittle in cold. A coating safe in dry halls may foul water after a flood.
Mark every experimental installation. State the material, date, expected life, inspection schedule, known reactions, and removal method. Workers arriving years later must be able to identify what they are drilling, heating, cutting, or washing away.
Do not conceal an experimental material behind finished surfaces until someone has accepted responsibility for inspecting it after the original research team leaves.
Public machines
A new automaton may sort cargo, direct passengers, inspect structures, clean streets, or carry supplies. Test it with damaged signs, crowded rooms, unusual bodies, children, frightened animals, broken cargo, poor lighting, conflicting instructions, and people who do not understand what the machine expects.
Consistency does not prove that an instruction is safe. A machine can repeat the same mistake perfectly.
Every public machine needs a visible stop method and workers with authority to use it. The machine's explanation of its own action should be preserved when possible, but the public must also be able to reach a person capable of changing the result.
If the machine shows memory, preference, judgment, fear, or an ability to object, the experiment may involve a participant rather than a tool. Stop treating its interior as laboratory property until that question has been examined.
People affected without volunteering
Some trials change streets, utilities, buildings, or transport used by people who never agreed to participate. Begin on the smallest workable scale. Announce the change in places people actually use. Provide a direct way to report faults and a clear signal that will stop the trial.
Do not dismiss a complaint because the instrument did not record the problem. The complaint may identify what the instrument was never placed to see.
Keep an unaffected route or manual method where possible. Restore the previous system quickly when the stop conditions are met. The ability to reverse a trial is part of the design, not an admission that failure is expected.
Liability
Set aside repair and compensation funds before the first public use. State who provides medical care, replaces damaged property, restores lost service, and pays workers whose shifts were canceled by the trial.
A person harmed by one failure should not have to prove that the entire invention is worthless. A useful system can still owe repair to the person it injured.
Publish failures as carefully as successes. Include abandoned trials, stop orders, near misses, unexpected maintenance, complaints, injuries, compensation, and changes made afterward. Future laboratories need to know which attractive ideas failed in rain, crowds, hunger, darkness, and ordinary hands.
A civic experiment belongs to the public because the public lives with the result. If the benefit is shared, the responsibility must be shared as well.
