*Technical transparency is insufficient if we ignore real-world effects. We must assess the impact, appeal mechanisms, and the capacity for choice of those affected by each algorithmic decision.*
The supplier's demonstration begins with a reassuring map. A municipal parks department can see coloured areas showing where visitors spend time: one shade suggests quiet use, another signals congestion, and a third marks a place that may need attention. The proposal is simple. Install occupancy sensors, combine their readings in a dashboard, and use the result to schedule cleaning, benches, repairs, or activities more efficiently. The technology is easy to imagine because it turns a familiar public space into a visible pattern.
At the end, a parents' association asks a different pair of questions. Will families and other residents feel safer in the park? Will the information actually change how the space is cared for? Those questions do not reject the device. They establish the outcome that would make the device worth considering. A park should support play, rest, meeting, and an unhurried journey home. Once that human result comes first, the sensor becomes one possible means rather than the reason for the project.
This order matters because what is easy to count can quietly become what the organisation values. A system may accurately record arrivals and dwell time while failing to notice a broken water fountain, a poorly lit path, a teenager avoiding harassment, or an older person who cannot find a seat in the shade. None of those experiences makes the data false. It shows that a partial signal cannot be allowed to speak as the whole condition of a place.
Begin with a change that people can recognise#
The first meeting should not decide where hardware goes. It should agree on a hypothesis of change. In this case, it might read: “If maintenance, orientation, and interventions respond to what people experience across the day, more residents will be able to use the park independently and stay there comfortably.” That sentence is deliberately modest. It does not claim that a dashboard produces safety. It says that public choices may improve daily use, and that the claim must be tested.
The first meeting should not decide where hardware goes. It should agree on a hypothesis of change.
A useful hypothesis exposes the links that sales language often hides. Information is gathered; someone interprets it; a team assigns work; an intervention changes the physical setting; people experience a consequence. Each link can fail. A sensor may work while a repair crew lacks time, a budget line may favour visible projects, or a busy corner may be misread because it is the only route to a school. Naming this sequence prevents an algorithm from becoming a convenient explanation for decisions that are really about staffing, priorities, or local knowledge.
It also makes room for alternatives. If faults take weeks to reach maintenance staff, a clear reporting channel and a reliable response route might solve more than an occupancy system. If an event is poorly attended because its timing conflicts with caring responsibilities, a conversation and a timetable change are more direct than another data feed. Comparing these options does not slow progress. It stops public money and public attention from being spent on an impressive answer to the wrong question.
Three human indicators before a procurement score#
The association proposes three indicators. The first is **practical access**. Can children, older people, and people with different mobility needs reach and use the areas intended for them? This means looking beyond attendance figures. It may involve walking the routes with a wheelchair user, observing a visit with a pram, and asking people where they pause or turn back. A full playground is not proof that everyone can use it; crowds can coexist with exclusion.
The second is **lived safety**. Formal incident reports are important, but they are not a complete account of whether people feel able to remain in a place. Many residents never report the insult, dark shortcut, or repeated behaviour that changes their routine. Conversations at different times, simple opportunities to describe avoided areas, and visible follow-up can reveal a problem before it becomes an emergency. Lived safety is not an unreliable extra beside “real” data. It is evidence about how lighting, maintenance, design, and social conditions reach people in everyday life.
The third is **the ability to influence a decision**. When someone notices an issue, do they know who is responsible? Can they find out what will happen next, and challenge a priority that seems mistaken? This indicator measures more than satisfaction. It tests whether the park remains a public space in which people have a meaningful route into public action. A system that speeds up internal work but makes explanations or complaints harder has increased administrative efficiency at the expense of civic agency.
Together, these indicators keep a sensor reading in proportion. High traffic in one corner may mean that more shade is needed; it may mean a path is functioning well; it may mean that every other entrance is inaccessible. Low traffic may reflect peace and quiet, or it may reveal a damaged play surface. A coloured map cannot settle that question. It needs to be checked against observation, repair records, and the knowledge of people who work in and use the space.
Make a pilot capable of failing#
The hypothesis should be allowed to be wrong. Before expanding a system, the council can run a time-limited pilot in one park, publish the purpose, and schedule a review after three months. Maintenance staff, local families, young people, older residents, and users who were not involved in the original design should all be able to inspect the result. The test is not “does the technology work?” That phrase is too vague to protect anyone. The test is whether accessible use, lived safety, and the ability to influence decisions improved without creating disproportionate harm.
An honest pilot records the inconvenient evidence as well. It notes signals that were interpreted incorrectly, reports that did not fit a category, people who chose not to share information, and the cost of keeping the system running. If the proposal needs cameras, identifiers, or visitor profiles to fulfil its promise, that requirement must be stated before deployment. The council must justify every element and show why a less intrusive option cannot meet the need. Being in a public park should not mean producing a permanent trail merely because an institution wants a better scheduling tool.
Clear responsibility is part of the design. One named role should answer for the public purpose, another for data practices, and a visible route should handle errors and review requests. A supplier can maintain equipment, but cannot be the sole authority on what counts as a successful public outcome. Nor should a silent software update be able to change thresholds, sources, or purposes. When the logic changes, people deserve a plain explanation and a real chance to question it.
Plurality makes the evaluation stronger. Cleaning staff understand routines that a map cannot display. Families see what changes after school. Young people may identify uses that adults misinterpret. Regular walkers often notice the small deterioration that reshapes a route long before it reaches a formal report. Listening to these accounts does not mean turning every preference into an automatic rule. It means explaining what evidence was received, what was decided, and why.
Choose the tool after the public purpose#
Only after this work should procurement begin. A sensor provider may still be the right choice, but the comparison will now be better. The council can ask what the system detects, how often it is wrong, where information travels, how long it remains, what happens during an outage, and how the service can be removed. A different option may be better: faster repairs, improved lighting, a staffed reporting channel, or regular neighbourhood meetings. Choosing the less glamorous measure is not a failure when it serves the agreed outcome more reliably.
Useful transparency is not a long technical document placed online after a contract is signed. It is the ability of a resident, municipal worker, or park user to answer ordinary questions: what are we trying to change, what information informs the decision, who is accountable, what are the limits, and how can an error be corrected? If only the people who bought the platform can understand those answers, the project has not yet earned public trust.
The next step can be small and concrete. Before requesting bids, the council writes the hypothesis of change and the three human indicators. It shares them with people who use the park at different hours, then explains which observations changed the plan. That practice moves power in a quiet but important way. The tool no longer dictates the question; it has to answer to a shared need. Good municipal technology is not the system that produces the most signals. It is the one that helps a common space become more usable and cared for without asking the people in it to give up their voice, privacy, or capacity to shape what happens there.
