We are becoming a society in which algorithms know more about our behavior than we do—and increasingly have the power to judge it.
We have spent the past two decades enthusiastically turning our lives into data. Our phones record where we go, our cars monitor how we drive, supermarkets track what we buy, employers measure our productivity, and social media companies analyze what captures our attention. Much of this happens without our knowledge, and almost all of it is justified in the name of convenience, efficiency, safety, or better service. We have gradually accepted that being watched is simply the price of participating in modern society.
Three articles illustrate where this arrangement may be taking us. In Ruled by Numbers: How Data Dominates Every Facet of Our Daily Lives, mathematician Noah Giansiracusa describes how ordinary human activities are increasingly reduced to numerical measurements. In an ACLU investigation into AI-driven police stops, Jay Stanley examines what happens when the same approach moves into law enforcement. And a remarkable investigation by The Colorado Sun shows how an innocent woman found herself accused of a crime largely because police placed too much confidence in surveillance records.
The connection is troubling. The technology that decides which advertisement you see can also help determine whether the police consider you suspicious.
Giansiracusa takes us through an ordinary day. We wake up and check our phones, where algorithms decide which information deserves our attention. We drive to the supermarket, generating information about our location, speed, braking habits, and movements. At the store, our purchases contribute to enormous databases used for marketing and business decisions. Back home, we apply for jobs, where automated systems may rank our qualifications before a human being ever reads our applications.
Even the workplace has become an exercise in measurement. Giansiracusa describes an investment firm that evaluates employees using 77 different measures. At a hospice organization, visiting a dying patient earns one productivity point, while attending a funeral earns 1.75. The absurdity is revealing. Activities involving judgment, compassion, and human relationships are converted into numbers because numbers are easier to compare, manage, and optimize.
There is nothing inherently wrong with measurement. Modern medicine, engineering, transportation, and economics would be impossible without it. Businesses need information to operate efficiently, and governments need statistics to make informed decisions. The trouble begins when measurement stops being a tool for understanding human activity and becomes a substitute for understanding human beings.
The ACLU article illustrates that danger particularly well.
In September 2025, a young woman named Grace Knowles was traveling across the United States when a police officer pulled her over. According to her account, the officer initially cited a minor traffic violation but later explained that a surveillance system had identified her travel patterns as suspicious. He showed her a computer display containing records of her movements, assembled from automated license plate readers positioned along American roads.
Knowles had not been accused of a serious crime. She was taking a road trip. Yet a system that had collected information about where her car had been apparently found something unusual enough to attract law enforcement attention. The computer had transformed ordinary travel into a reason for suspicion.
The Colorado case demonstrates an even more immediate danger.
In September 2025, Chrisanna Elser, a Denver financial adviser, received an unexpected visit from a police sergeant who accused her of stealing a package worth no more than $25 from a neighboring community. The officer was confident he had identified the culprit. Automated license plate cameras operated by Flock Safety had recorded her green Rivian traveling through the area around the time of the theft. The officer also claimed that doorbell footage supported the accusation.
He arrived at her home with a summons and insisted that the evidence was conclusive. According to doorbell footage reviewed by The Colorado Sun, he told her there was no doubt about her guilt.
But there was a problem. Elser had not stolen anything.
She had driven through the neighborhood that day to visit her tailor, more than a quarter-mile from the scene of the theft. To establish her innocence, she spent days assembling evidence from her phone’s location history, her vehicle’s GPS records, dashboard cameras, surveillance footage from the tailor, and statements from people she had encountered.
The police eventually dismissed the summons after she submitted her evidence. But the episode had already consumed considerable time, caused significant anxiety, and raised concerns about her professional reputation. Elser had effectively been forced to conduct her own investigation to disprove an accusation that the police had not adequately verified.
What makes this case especially revealing is that the surveillance system did not necessarily provide inaccurate information. Elser really had driven through the neighborhood. The mistake arose when police treated that information as evidence of something it could not establish: that she had committed a theft.
This is the distinction that too often gets lost in discussions about artificial intelligence. Data can be accurate while the conclusions drawn from it are completely wrong. A license plate camera can establish that a vehicle passed a particular location at a particular time. It cannot establish why the driver was there, what the driver intended, or whether the driver committed a crime.
Yet the apparent precision of computerized information can encourage unjustified confidence. A timestamp, photograph, and database record look authoritative. When several such records appear together on a screen, they can create the impression of a complete and convincing account of events. In reality, they may provide only fragments of a much larger story.
The Colorado case also exposes a disturbing reversal of responsibility. In our legal system, the government bears the burden of proving criminal guilt. In Elser’s experience, the practical burden fell on her to prove that the government’s accusation was mistaken. Fortunately, she had the technology, resources, and determination to do so. Not everyone would.
And here the three articles come together.
Commercial data collection and government surveillance are not identical, and their consequences differ enormously. A supermarket using purchasing history to offer discounts is not the same as a police department using travel records to identify suspects. But both depend on the assumption that human behavior can be recorded, analyzed, categorized, and predicted through mathematical models.
Artificial intelligence makes this process enormously more powerful. It can examine millions of records, discover patterns that human analysts might overlook, and identify relationships between seemingly unrelated events. These capabilities can be valuable in detecting fraud, finding missing persons, identifying stolen vehicles, and investigating serious crimes.
But discovering a statistical relationship is not the same as establishing guilt, intent, or reasonable suspicion. An unusual travel pattern is not evidence of a crime simply because a sophisticated computer has identified it. And being near the scene of a crime does not make someone the criminal.
This is where the Fourth Amendment becomes critically important.
The Fourth Amendment protects Americans against unreasonable searches and seizures. Its purpose is to place limits on government power, particularly the ability of authorities to intrude into private lives without adequate legal justification. Those protections cannot be allowed to become meaningless simply because surveillance has become automated.
Police departments must become far more sophisticated in their handling of artificial intelligence, or the protections afforded by the Fourth Amendment risk being steadily eroded. A computer-generated alert should not become a convenient shortcut around constitutional requirements. An algorithm identifying suspicious behavior does not automatically establish reasonable suspicion for a detention or probable cause for a search.
There is an important legal distinction here. Courts have generally permitted police to observe and record license plates visible on public roads. But the constitutional implications of collecting vast amounts of location information, retaining it, and using AI to reconstruct an individual’s movements are far more complicated. The Supreme Court has already recognized, in other contexts involving prolonged electronic location tracking, that the accumulation of information can raise constitutional concerns beyond those presented by isolated observations.
The danger is that practices which appear lawful when considered individually may produce something fundamentally different when combined. One camera recording a passing automobile is ordinary observation. Thousands of cameras documenting a person’s movements over months, combined with AI systems searching for patterns and associations, amount to a much more intrusive form of surveillance.
The responsibility rests squarely with law enforcement leadership. Police departments need clear policies, proper training, independent audits, and strict limits on when AI-generated information may justify official action. Officers must understand what these systems can establish, what they cannot establish, and how easily accurate data can produce false conclusions.
Equally important, officers must be willing to question the technology rather than defer to it. The Colorado case illustrates what happens when an investigator becomes convinced that the available electronic records tell the entire story. Instead of considering alternative explanations, the officer reportedly dismissed the suspect’s attempts to present contrary evidence.
Courts and legislatures also have a responsibility to ensure that constitutional protections remain meaningful as surveillance capabilities expand. The question is not whether police should be allowed to use modern technology. Of course they should. The question is whether technological capabilities should be permitted to change the practical meaning of constitutional rights without adequate public debate, legislation, or judicial oversight.
There is another danger that deserves attention: people may begin changing their behavior because they know they are being watched. Someone who believes every journey is recorded and analyzed might think twice about attending a political demonstration, visiting a controversial organization, or simply traveling somewhere that could attract unwanted attention. No law needs to prohibit these activities. The possibility of scrutiny can be enough to discourage them.
This is particularly troubling in a democratic society, where freedom means more than the absence of formal restrictions. It also means being able to conduct ordinary, lawful activities without feeling obliged to explain them to the authorities.
None of this requires abandoning technology or pretending that society would function better without data. Artificial intelligence offers extraordinary opportunities to improve public safety and government efficiency. But those benefits depend on institutions understanding the technology’s limitations and accepting responsibility for the decisions made with its assistance.
Giansiracusa argues that ordinary people need a better understanding of the mathematical systems shaping their lives. That is a sensible starting point. But mathematical literacy alone cannot solve a problem involving unequal power. Citizens also need enforceable rights, independent oversight, and institutions willing to recognize that technological capability does not automatically establish legal or moral justification.
For centuries, governments and businesses have tried to measure and classify the people they govern, employ, and serve. What has changed is the scale, speed, and intimacy of that process. Information that once required substantial effort to collect can now be gathered automatically, stored indefinitely, and analyzed almost instantaneously.
The real question, therefore, is not whether we should use artificial intelligence to understand human behavior. We already do, and much of that work is beneficial. The question is how much authority we are prepared to give the conclusions it produces.
We have created extraordinarily powerful tools for turning human lives into numbers. We must now make sure that those numbers do not become more important than the people they describe.
The Fourth Amendment was written to protect citizens from unreasonable government intrusion, not merely from the limitations of eighteenth-century investigative technology. Its principles must remain just as effective in the age of artificial intelligence.
After all, there is a considerable difference between a computer knowing where you have been and a computer deciding that you should have to explain why.
