When Code Meets the Constitution

In January 2021, a forty-three-year-old man in Grand Rapids, Michigan, received a letter that would upend his life. The Michigan Unemployment Insurance Agency had determined—based on evidence the agency refused to disclose—that he had committed fraud. He owed the state $7,200. Interest was accruing. If he did not repay the funds within thirty days, his wages would be garnished, his tax refunds seized, and his case would be referred to the Department of Justice for potential criminal prosecution.

The problem, as he would eventually learn, was that he had never filed an unemployment claim. Someone else had used his identity—their name, Social Security number, employment history—to file a fraudulent claim during the pandemic chaos of 2020. The Michigan unemployment system, overwhelmed by a tsunami of legitimate claims, had deployed an automated fraud detection algorithm called MIDAS (Michigan Integrated Data Automated System) to sort the genuine from the bogus. MIDAS had flagged his identity as suspicious. The system had generated a fraud determination without human review. And the system had issued that determination automatically, without ever informing the actual human being whose life it was about to destroy.

For eight months, this man—let's call him Robert, since he requested anonymity out of fear that publicly discussing his case would somehow make things worse—tried to clear his name. He called the unemployment agency. He waited on hold for six, seven, eight hours at a time. He submitted identity documents. He wrote letters. He filed appeals. Each time, he was told that the system's determination was based on "algorithmic analysis" and that the specific reasons for his flagging could not be disclosed because revealing the methodology would "compromise fraud detection capabilities."

Robert was not alone. During the pandemic, Michigan's MIDAS system falsely flagged more than 40,000 people for fraud. Thousands of others faced similar algorithmic accusations across multiple states. Some people had their bank accounts frozen. Others had wages garnished. A few—the most desperate, the most marginalized—lost their modest apartments because they couldn't pay rent while waiting for their frozen benefits to be restored. One woman in Detroit told a reporter that she had considered suicide. Not because she had done anything wrong, but because an algorithm had decided she was a criminal, and there seemed to be no way to prove otherwise.

This is the story at the heart of the collision between code and constitution. It is a story about what happens when the ancient promises of due process—your right to know what is being alleged against you, your right to confront your accusers, your right to be heard before you are condemned—run into the cold efficiency of automated decision-making. It is a story about power, accountability, and the question of whether the institutions we built to protect individual rights in the age of bureaucracy are adequate to protect those same rights in the age of algorithms.

To understand how we arrived at this moment, we need to travel back through history—not to the invention of the computer, but to an earlier revolution in administration that gave us the tools we still use today to check bureaucratic power. Then we need to trace the path from those earlier administrative systems to the computational networks that now govern so much of our lives. And then we need to grapple with what it means that the machinery of judgment—the mechanisms by which governments decide who is entitled to benefits, who should be investigated, who deserves a chance—has been transferred from human beings with faces and names to machines with processes and parameters.

The Administrative Revolution and the Origins of Due Process

The story of how modern governments came to terms with bureaucratic power begins not in Silicon Valley but in nineteenth-century industrial cities, where the growth of the modern state created new forms of authority that the founders of democratic republics had never anticipated.

When the framers of the American Constitution drafted the Due Process Clause, they were concerned primarily with the power of kings. The Fifth Amendment's prohibition against deprivation of "life, liberty, or property, without due process of law" was a response to the arbitrary exercise of monarchical authority—the ability of a sovereign to dispossess a subject, imprison a dissenter, or confiscate property on a whim. The solution they devised was procedural: before the government could act against an individual, it had to follow certain forms. There would be notice. There would be an opportunity to be heard. There would be an impartial tribunal. These requirements did not guarantee correct outcomes, but they did guarantee that outcomes could not be reached arbitrarily.

For roughly a century, this framework worked reasonably well. Government was small enough that most interactions between state and citizen remained personal. If the tax collector claimed you owed money, you could go to court and challenge his calculation. If the sheriff arrested you for a crime, you could demand to see the evidence against you. The machinery of state power was visible, accountable, and subject to human intervention.

Then came the administrative state.

The late nineteenth and early twentieth centuries witnessed an explosion in the scale and scope of governmental activity. The Industrial Revolution had created economic entities—large corporations, national railroads, vast financial institutions—whose operations transcended local boundaries and exceeded the comprehension of traditional legal institutions. The Progressives who rose to power in response to these changes believed that specialized agencies, staffed by experts, could manage these new complexities more effectively than courts bound by rigid procedural rules.

They were right, in a sense. The new administrative agencies could regulate railroads, oversee food safety, manage unemployment insurance, and administer welfare benefits with an efficiency that traditional institutions could never match. But they also concentrated enormous power in the hands of unelected officials operating outside the normal framework of judicial oversight.

The legal scholar Ernst Freund, writing in 1894, captured the dilemma that Progressives faced: "The great problem of administrative law is to secure the advantages of a flexible and efficient organization without surrendering the essential guarantees of individual liberty." The solution they devised—a compromise between efficiency and rights—was the modern administrative hearing. When the government wanted to take action that would affect an individual's interests, it would provide notice, allow the affected party to present evidence, permit cross-examination of adverse witnesses, and issue a written decision explaining the agency's reasoning. These procedures were less formal than full judicial proceedings, but they preserved the essential elements of due process: notice, hearing, and reasoned decision.

This framework served reasonably well throughout most of the twentieth century. Yes, there were abuses. Yes, some administrative agencies became captured by the industries they were supposed to regulate. Yes, the sheer volume of bureaucratic decisions meant that many individual cases received less attention than they deserved. But the system had teeth. Citizens could appeal adverse determinations. Courts could review agency actions for arbitrariness. And the administrative law judge—a creature of the mid-twentieth century, trained in both legal procedure and the substantive expertise required by the agency—served as a crucial check on raw bureaucratic power.

The system had another feature that is easy to overlook: it was legible. When a bureaucrat denied your benefits, you knew why. When an agency investigator recommended an enforcement action, that recommendation had to be do