In October 2013 Michigan’s Unemployment Insurance Agency fully implemented the Michigan Integrated Data Automated System (MiDAS), a roughly 47 million dollar replacement for a decades-old mainframe. As the state Auditor General’s February 2016 performance audit records, MiDAS was meant to collect unemployment taxes and pay benefits, with goals including increased automation and improved data accuracy; in fiscal year 2014 the agency paid one billion dollars in benefits across more than 611,000 claims.
In operation, MiDAS produced a flood of false fraud determinations. The system was configured to treat data discrepancies as evidence of intentional fraud with little or no human review, and it used income averaging that misstated when claimants had earned money. People received automated fraud findings and were assessed penalties up to four times the disputed amount, often without understanding why. Independent reviews later concluded that the large majority of the agency’s automated fraud determinations during this period were wrong, affecting tens of thousands of residents.
The Auditor General’s audit found material weaknesses in the security and access controls around MiDAS and the unemployment data it held. Litigation followed for years; the agency agreed to stop using the system’s automated fraud functions without human review, and in 2022 the parties settled a major class action for 20 million dollars.
MiDAS is a clear warning that automating a high-stakes accusation - here, fraud - without human judgment and without a workable way to contest the result turns ordinary administrative errors into life-altering harm. Speed and cost savings counted for nothing against the damage of being wrongly branded a criminal by a computer.