> ## Content Index > Fetch the complete content index at: https://www.nanosava.com/llms.txt > Use this file to discover other available public pages before exploring further. # Part II - Our New, Global Panopticon - URL: https://www.nanosava.com/essays/part-ii-our-new-global-panopticon/ - Published: 2026-02-21T00:01:00.000Z - Updated: 2026-10-05T13:48:03.000Z - Description: The Automaton of Justice — Algorithmic Determinism and the Death of Discretion - Author: Sava - Tags: Panopticon *In Part I, we traced the architecture of total surveillance—from Richelieu’s six lines to the 181 zettabytes of data humanity now generates each year. The archive exists. The question is: who reads it, and what do they conclude? The answer, increasingly, is not a person at all.* ## 2.1 The Reversal of Investigation: From Suspect to Data The transition from human-led investigation to algorithmic policing represents a fundamental inversion of the justice system. Traditionally, an investigation began with a crime and worked toward a suspect through evidence. Algorithmic policing works in reverse: it begins with a massive dataset of “innocent” activity and uses algorithms to generate suspects based on correlation and pattern matching. This shift has birthed a new category of injustice—the “false positive” life—where citizens are compelled to prove their innocence against the opaque determination of a black-box system. The Richelieu Wager has been automated. The Cardinal’s genius lay in his willingness to scrutinize the innocent until guilt appeared; the algorithm does this at scale, without malice, without discretion, and without the possibility of looking into a suspect’s eyes and recognizing a mistake. ## 2.2 The Terror of the “Hot List”: Automated License Plate Readers Automated License Plate Readers (ALPRs) are the sensory organs of the dragnet. These high-speed cameras, mounted on police cruisers, streetlights, and highway overpasses, capture thousands of license plates per minute, converting images into machine-readable text and cross-referencing them against “hot lists” of stolen vehicles, warrant subjects, and other persons of interest.²² While the public justification is often the recovery of stolen cars, the reality is a system of mass surveillance that logs the movements of millions of non-criminal drivers. The vast majority of scans—often exceeding 99 percent—belong to innocent motorists.²⁴ This data is pooled into regional databases and retained for months or years, creating a comprehensive map of a population’s movements. In states like California, data sharing is restricted, but in many jurisdictions, retention periods are effectively indefinite.²² The danger of ALPRs lies not only in privacy invasion but in what researchers call “automation bias”—the tendency of law enforcement officers to trust the machine over their own observation. ALPR systems are prone to errors, including Optical Character Recognition failures (misreading a “D” as an “O”) and “database hygiene” issues such as failing to remove recovered vehicles from the hot list.²⁵ ### Case Study: Aurora, Colorado (2020) The terrifying potential of ALPR error was realized in Aurora, Colorado, when police pulled over a Black family in an SUV. The system had flagged their license plate as a match for a stolen motorcycle from Montana. Despite the vehicle being a car with Colorado plates—a clear visual mismatch—the officers, conditioned to trust the “hit,” conducted a high-risk felony stop. They forced the mother and her children, including a six-year-old, to lie face-down on the hot pavement at gunpoint.²⁶ The machine’s error became the officers’ reality, bypassing basic visual verification in favor of algorithmic certainty. Richelieu needed six lines to hang a man. The ALPR needed one misread character. ### Case Study: The “Fog Light” Error (Detroit) In Detroit, police used ALPR data to identify all Dodge Chargers in the vicinity of a shooting. Isoke Robinson’s vehicle was flagged two miles from the crime scene. The suspect vehicle was known to have a missing fog light; Robinson’s car did not. Despite this exculpatory physical evidence, the “data match” of the license plate led officers to arrest her, impound her car, and place her toddler in a patrol vehicle.²⁵ Dragnet logic—identifying all vehicles of a certain type across a broad area—had replaced individualized suspicion with statistical probability. ### Case Study: Denise Green (San Francisco) Denise Green was subjected to a similar trauma when an ALPR system misidentified her vehicle as stolen. She was pulled from her car at gunpoint—a confrontation that could easily have escalated to lethal violence.²⁷ These cases demonstrate that ALPRs do not merely “read” plates; they actively construct threat assessments upon which officers act with lethal intensity. ## 2.3 The Face of the Wrong Man: Facial Recognition Failures If ALPRs track our possessions, Facial Recognition Technology (FRT) tracks our biological identity. The deployment of FRT in policing has been fraught with controversy due to its documented racial bias and high error rates, particularly when identifying people of color.²⁸ ### Case Study: Robert Williams (Detroit) The case of Robert Williams is a landmark example of algorithmic wrongful arrest. In January 2020, Williams was arrested in his driveway in Farmington Hills, Michigan, in front of his wife and young daughters. The Detroit Police Department had run grainy surveillance footage of a shoplifter through a facial recognition system, which matched the suspect to Williams’ driver’s license photo. No other evidence linked him to the crime. He had never visited the store in question. Williams was held in a crowded detention center for 30 hours. When detectives finally showed him the surveillance photo, he immediately pointed out that the man in the image was not him. The detective, confronted with the discrepancy, remarked that “the computer must have gotten it wrong.”²⁸ The arrest was purely the result of an algorithm’s confidence score being treated as probable cause. ### Case Study: Michael Oliver (Detroit) Another Detroit man, Michael Oliver, was accused of felony larceny after an FRT match identified him as the perpetrator in a video of someone reaching into a truck. The actual perpetrator had different tattoos and arms than Oliver. The reliance on the facial match blinded investigators to obvious physical discrepancies that a human observer should have caught.²⁸ These cases expose the “black box” nature of algorithmic justice. The defendants were not identified by witnesses or forensic evidence, but by proprietary software matching pixels. The “computer says guilty” dynamic places the burden on the accused to prove a negative—often after the trauma of arrest and incarceration has already occurred. ## 2.4 The Reverse Warrant: Geofence Dragnets and the Fishing Expedition The geofence warrant may be the most invasive digital investigative tool currently in use. Unlike a traditional warrant, which targets a known suspect at a specific location, a geofence warrant targets everyone within a specific area at a specific time. It works backward—from the crime scene to finding a suspect.³⁰ The process typically unfolds in three steps. First, police order Google—the primary holder of such data via its Location History feature—to release anonymized IDs of all devices within a defined geographic perimeter during the timeframe of a crime. Second, investigators analyze the movement patterns of these anonymous IDs to identify “suspicious” behavior. Third, police request the subscriber information—name, email address—for a narrowed-down list of IDs.³⁰ ### Case Study: Zachary McCoy (Gainesville, Florida) Zachary McCoy, an avid cyclist, used the RunKeeper app to track his rides. This app fed data into Google’s Location History. When a burglary occurred in his neighborhood, McCoy’s bike route—which involved looping around the area—placed him near the burglarized home three times within an hour. This data pattern, innocuous to McCoy but suspicious to an algorithm, flagged him as a prime suspect in a geofence warrant. McCoy was fortunate: Google notified him of the warrant request, allowing him to hire a lawyer and successfully block the release of his personal data. Had he not received that notice, he likely would have been arrested solely on the basis of his exercise habits.³¹ ### Case Study: Jorge Molina (Avondale, Arizona) Jorge Molina was not as fortunate. Police investigating a murder used a geofence warrant that placed a device associated with Molina’s Google account at the scene. He was arrested, held in jail for six days, and publicly named as a murder suspect. He lost his job, his car, and dropped out of school. Later investigation revealed that Molina was innocent; the location data was likely triggered by a device he had once logged into—perhaps a relative’s phone—or was a result of GPS drift.³² The “certainty” of the GPS coordinate destroyed his life before the error could be rectified. ### Legal Battleground: The Circuit Split The constitutionality of geofence warrants is currently the subject of intense judicial debate. In 2024, the U.S. Court of Appeals for the Fifth Circuit (United States v. Smith) ruled that geofence warrants are unconstitutional “general warrants,” finding that users have a reasonable expectation of privacy in their location history. Conversely, the Fourth Circuit (United States v. Chatrie) ruled that users essentially “opt in” to location tracking, thereby waiving their privacy rights under the Third-Party Doctrine.³⁴ This circuit split creates a fragmented legal landscape in which a citizen’s protection from dragnet surveillance depends entirely on their geographic location—an irony not lost on privacy advocates. --- *The cases above document what happens when the algorithm reaches into individual lives—the Aurora family, Robert Williams, Zachary McCoy, Jorge Molina. They are the visible victims. But for every person caught in the dragnet’s direct path, millions more are altered invisibly. In Part III, we examine the subtler, more pervasive harm—the psychological transformation of an entire society that knows it is being watched.* ## Works Cited 22\. “Summary: Automated License Plate Readers — State Statutes,” NCSL, accessed January 30, 2026 24\. “REPORT: ACLU releases comprehensive report on law enforcement’s use of license plate readers,” ACLU of SoCal, accessed January 30, 2026 25\. “The Human Toll of ALPR Errors,” Electronic Frontier Foundation, accessed January 30, 2026 26\. “When license plate readers get it wrong,” CBS News, accessed January 30, 2026 27\. “San Francisco Woman Pulled Out of Car at Gunpoint Because of License Plate Reader Error,” ACLU, accessed January 30, 2026 28\. “Williams v. City of Detroit,” American Civil Liberties Union, accessed January 30, 2026 29\. “Flawed Facial Recognition Technology Leads to Wrongful Arrest and Historic Settlement,” Quadrangle, Michigan Law, accessed January 30, 2026 30\. “Do Geofence Warrants Violate the Fourth Amendment?” Lawfare, accessed January 30, 2026 31\. “Google data puts innocent man at the scene of a crime,” Sophos, accessed January 30, 2026 32\. “Geofence Warrants and the Fourth Amendment,” Harvard Law Review, accessed January 30, 2026 34\. “Much Ado About Geofence Warrants,” Harvard Law Review Blog, accessed January 30, 2026