Every city I've lived or worked near has slowly filled up with cameras on streetlights, in matatus, above shop doors, inside apps that ask for your location before they'll let you see the menu. Each one arrives selling the same fantasy: enough eyes in enough places, and the whole messy problem of human wrongdoing conveniently resolves itself, a fully monitored world sold as a kind of utopia, order achieved without anyone having to do the harder work of becoming good. It is a promise that mostly delivers, in the narrow sense it's sold on, behavior does change when people know they're being observed. But something else changes with it, more slowly and far less visibly: the muscle a person uses to ask why is this wrong, rather than will I get caught, gets less exercise every year a camera does that job instead.
The citizen surveillance is built to produce
Surveillance doesn't just watch a population, over enough time, it shapes one. Not through persuasion, but through selection pressure: the behaviors that survive are the ones a lens can see and a rule can name. Loitering, trespassing, unlicensed selling, a face that lingers too long near the wrong entrance all trivially detectable, all enforceable at scale. And because the system rewards exactly this (staying inside the frame, looking unremarkable, generating no flag), the citizen it produces learns to treat compliance itself as the finish line. Why a father is loitering outside a school gate, why a vendor has no license, why a neighborhood keeps producing the exact young men an algorithm flags as "high-risk": these become questions the compliant citizen is never made to ask. Not because the answers are hidden from them, but because the architecture of control has already absorbed the consequence on their behalf, so the habit of asking why atrophies from simple disuse. They end up outwardly compliant, socially palatable, and behaviorally acceptable, yet incurious, in the one respect that matters most, about the conditions inside their own community that keep producing the thing being policed. The camera's blind spot becomes their blind spot too, inherited rather than chosen.
Morality reduced to enforcement
There is a difference between a person who doesn't steal because they've reckoned with what theft does to a neighbor, and a person who doesn't steal because a camera is pointed at the till. Both produce the same outcome on the security report. Only one produces a person who can be trusted in the room the camera doesn't reach. Surveillance, by design, collapses that distinction: it treats the absence of an observed violation as evidence of virtue, when it is at most evidence of visibility. A society that measures its own goodness this way is measuring the wrong thing, and it will keep getting the wrong answer back: crime statistics that look better while trust between neighbors, the kind that doesn't show up in any dashboard, keeps eroding.
That substitution has a cost past the individual, too. Enforcement can suppress a behavior without anyone ever having to name what produced it. The structural conditions behind it (unemployment, a transit system that strands people exactly where the cameras are densest, housing policy that concentrates the people being watched) and the cultural ones (which harms a community has tacitly agreed to tolerate, whose complaints get taken seriously and whose don't, what "normal" has been allowed to mean) both get to stay unexamined, because the dashboard already shows the numbers improving. Confronting either is slow, political, and uncomfortable. Adding another camera is none of those things. Institutions will reliably choose the second when the first is on offer, and every time they do, whatever was actually breeding the wrongdoing gets to keep breeding it, just further from view.
This is the part that should trouble anyone who builds the systems doing the watching, and I count myself in that group. Most of the infrastructure I work with (anomaly detection, pattern recognition over behavioral data, models trained to flag "unusual" activity) is architecturally indifferent to the question of why the pattern exists. It is built to notice deviation, not to understand cause. That's a reasonable scope for a fraud model. It is a dangerous scope for a society's entire moral apparatus, and it is exactly the scope surveillance silently assumes once it moves from catching specific crimes to managing "public safety" in general. The tool was never wrong. It was just asked to answer a question it was never built to answer.
Where the evil actually goes
None of this means surveillance fails to suppress wrongdoing. It visibly does, in the spaces it covers. The failure is in what happens to whatever it suppresses. Evil, in the plain sense of the word (exploitation, cruelty, corruption, the deliberate harming of people who can't retaliate), doesn't need a camera-free alley to survive. It needs only a gap between what is watched and what is understood, and every surveillance regime, however dense, has one. Domestic abuse rarely happens on a monitored street. Corruption moves through channels built specifically to look unremarkable to whatever is watching for it. The most sophisticated harms in any surveilled society are, almost by definition, the ones that have already adapted to being watched. What gets caught is the crude and the desperate. What gets driven underground is patient enough to wait the camera out, and it comes back smarter about where the blind spots are.
The good person the system produces
The sharper danger doesn't generate an incident report, which is exactly why it's easy to miss. It's not the criminal the system catches. It's the ordinary, outwardly decent person the system produces: someone whose entire moral architecture was built to satisfy an external observer, and who has therefore never had to build one of their own. Take the observer away: move them to a country with no cameras, put them in a position of unmonitored power, hand them a colleague's mistake nobody else will ever see, and there is no internal structure left standing to catch what the external one used to catch. Conscience is not a passive good record. It's a trained capacity to notice harm and feel the pull to stop it even when nothing external is checking. A person raised inside enough surveillance to never need that capacity will look, by every visible measure, exactly as good as someone who built it the hard way, right up until the one moment that actually requires it.
What a camera cannot substitute for
None of this is an argument against every camera, every audit trail, every fraud model. Plenty of harm really is deterred by the simple fact of being observable, and I'm not interested in romanticizing a version of society with no accountability infrastructure at all. The argument is narrower: observation is not a substitute for formation, and a society that casually lets it become one is trading a resource it can't easily rebuild once it's gone. Conscience gets built the way any capacity does: through practice, through being trusted with situations where the right thing and the easy thing diverge and no one is checking which one you picked. Every system that removes that divergence, in the name of safety, is also removing the training ground. The honest question for any institution reaching for a surveillance solution isn't just will this catch more violations. It's what capacity are we no longer asking people to build, now that this is watching for us. A society that keeps answering only the first question, never the second, won't end up more virtuous for all its watching. It will end up thoroughly monitored, outwardly orderly, and quietly unable to see, in itself, the thing it built the cameras to find.