Michigan’s MiDAS Unemployment Algorithm Scandal

Michigan’s Unemployment Insurance Agency used the Michigan Integrated Data Automated System (MiDAS) to process claims and identify discrepancies that could suggest fraud. The important distinction is between a system identifying a discrepancy and a government decision that a person intentionally misrepresented information. A mismatch can arise from timing, payroll reporting, an employer correction, or a misunderstood question. Treating every mismatch as fraud can trigger repayment demands, penalties, or collection actions without a fair chance to resolve the facts. The Michigan Office of the Auditor General audited MiDAS information-processing controls in 2016. That audit had a stated scope and explicitly excluded review of the system’s fraud identification, which was examined separately. Michigan’s Attorney General later announced a $20 million settlement resolving a civil-rights class action that alleged the UIA used an automated system to falsely accuse claimants of fraud and seize property without due process. The announcement describes settlement of allegations, not a court trial finding that every determination was false. State officials also reported refunds after reviews of prior fraud determinations. The system became a prominent case study because automation changed both the scale and speed of administrative decisions. When an automated process produces notices, claimants need enough information to understand the specific issue: which week, employer, income record, rule, and calculation are involved. A generic notice or complex online portal can make it difficult to contest a wrong record. Staff should examine the source evidence, let the claimant respond, and distinguish suspected error from intentional fraud. Reliable safeguards include human review before a fraud finding, clear and timely notice, accessible appeal channels, audit logs, and monitoring of reversal rates and demographic or language barriers. Agencies should test data matching with real edge cases and examine how the system handles late or corrected wage reports. Vendors should support audits and explainable records. The MiDAS episode does not show that every automated benefit tool causes harm.