The Trial Nobody Sees: When Algorithms Judge

The Trial Nobody Sees: When Algorithms Judge

The Harrison Family's Nightmare

On a sweltering Tuesday in July 2019, the Harrison family of Dayton, Ohio, experienced what could only be described as a modern nightmare—not one involving monsters under beds or intruders in the night, but something far more insidious: a faceless system that had decided they were criminals without ever looking them in the eye.

Linda Harrison, a 47-year-old school cafeteria worker, had just returned home from her evening shift when she noticed the lights were out. Not the normal darkness of an empty house, but a deliberate, enforced blackness. The power company had cut them off. When she called the emergency number, a recorded voice informed her that service had been terminated due to "confirmed evidence of theft and meter tampering"—a Class A utility felony under Ohio law.

"We'd been customers for eleven years," Linda would later tell investigators. "We never missed a payment. We never touched that meter. But the computer said we did, and there was no one to argue with."

What Linda Harrison didn't know—what she had no way of knowing—was that she had just become a defendant in a trial unlike any other in American history. The judge was an algorithm. The jury was a neural network. The evidence was a probability score generated by a system that her utility company had purchased from a third-party vendor, a system whose inner workings were protected by trade secret law and whose decisions could not be appealed to any human court.

Over the next seventeen months, the Harrisons would discover just how destructive an invisible trial could be. The power shutoff triggered a cascade of consequences: their lease agreement was voided when the landlord learned of the "theft" accusation, forcing them to find new housing during a pandemic. Linda's employer, a school district with strict policies around employee integrity, placed her on administrative leave pending an investigation that would take four months to resolve—four months without pay while she cared for her two children, one of whom had severe asthma that required a working nebulizer. Her husband Marcus, a freelance electrician, saw his client base evaporate when word of the accusation spread through their small community.

The most infuriating part, Linda would repeatedly emphasize in interviews, was that there was no court to appeal to. No judge to explain herself to. No jury to persuade. When she finally managed to reach a human representative at the power company, she was told the decision was "system-generated and cannot be overridden." When she sought legal counsel, three attorneys told her the same thing: the algorithm's determination was considered a "business decision" rather than a legal judgment, placing it outside the reach of consumer protection law.

It would take eighteen months, a complaint to the state utility commission, a news article that went viral on social media, and intervention by a civil liberties organization before the Harrisons' power was finally restored, their name cleared, and a written apology issued. By then, the damage to their credit, their housing stability, and their mental health was largely irreversible.

"The worst part," Linda Harrison said in her testimony before a state legislative committee, "wasn't the lights going out. It was knowing that something had decided we were guilty without ever giving us a chance to explain. We didn't even know we were on trial until the verdict came down."

---

The Harrison family's experience, while extraordinary in its cruelty, is far from unique. Across the United States and throughout the developed world, ordinary citizens are finding themselves increasingly subject to decisions made by systems they cannot see, cannot question, and often cannot even access. These algorithmic decision-makers have infiltrated every corner of daily life, from the credit limits assigned to our bank accounts to the prices we're charged for insurance, from the job applications that get forwarded to hiring managers to the bail amounts set for defendants awaiting trial.

We have, without quite noticing, created a shadow justice system—one that operates parallel to the courts we learned about in civics class, one that is often more consequential than the formal legal system, and one that is almost entirely invisible to the citizens it judges.

Consider what happens when you apply for a mortgage. Before a human loan officer ever glances at your application, an automated underwriting system has already processed your data, assigned you a risk score, and made a preliminary determination about your creditworthiness. The decision that seems to come from your local bank is often actually coming from an algorithm developed by Fannie Mae or Freddie Mac, or from a private vendor like Black Knight Financial Services. These systems evaluate thousands of data points—your payment history, your debt-to-income ratio, the racial composition of your neighborhood, even the make and model of car you drive—and reduce you to a single number. That number determines not just whether you get a loan, but at what interest rate, and ultimately what kind of house you can afford in what kind of neighborhood.

Or consider what happens when you're arrested. In jurisdictions across America, algorithms now help determine bail amounts—systems like PSA (Public Safety Assessment) developed by the Arnold Foundation or COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) developed by Equivant. These tools claim to predict your likelihood of reoffending, and their recommendations can mean the difference between going home to your family that night or spending months in a jail cell awaiting trial. A 2016 investigation by ProPublica found that COMPAS was nearly twice as likely to falsely flag Black defendants as future criminals compared to white defendants. Despite this finding, and subsequent academic criticism of the system's methodology, COMPAS continues to be used in courts across the country.

The pattern repeats in housing, in healthcare, in education, in employment. Algorithmic systems now make determinations that once required human judgment, human compassion, and human accountability. And while these systems are often presented as neutral, efficient, and objective—the perfect bureaucratic servants—the reality is far more troubling.

---

The Rise of the Algorithmic State

To understand how we arrived at this moment, we need to trace the history of automation in governance—a history that stretches back further than most people realize but has accelerated dramatically in recent decades.

The story begins, appropriately enough, with the IRS. In 1962, the Internal Revenue Service deployed one of the first large-scale automated decision-making systems to flag potential tax fraud. The system was crude by modern standards—essentially a series of rule-based filters that flagged returns matching certain patterns—but it demonstrated the potential power of algorithmic governance. Within a few years, the IRS was auditing hundreds of thousands of taxpayers based entirely on automated determinations.

What the IRS learned, and what other government agencies soon discovered, was that algorithmic systems offered something human bureaucracy could not: consistency, scalability, and deniability. A human auditor might be influenced by a taxpayer's demeanor, by implicit biases, by simple exhaustion. An algorithm, the theory went, would apply the same rules to everyone, regardless of race, gender, or social status. The algorithm didn't have a bad day. The algorithm didn't hold grudges. The algorithm didn't take bribes.

These were seductive promises, and they drove adoption across government agencies. The Social Security Administration began using automated systems to flag potential fraud in benefit claims. State welfare departments deployed algorithms to detect overpayments and improper e