evidoria

← Back to browse

Good practice Imported

MiDAS — Michigan's Automated Unemployment Fraud Algorithm That Wrongly Accused Tens of Thousands

United States of America · Lansing · See the United States of America profile

Evidence: Descriptive / self-reported Top 97% 20/100 · Ask Evidence Copilot about this practice

Michigan's MiDAS system auto-adjudicated unemployment fraud claims with no human review (2013-2015), wrongly accusing 34,000+ people at an ~85% error rate; a $20M settlement followed in 2022.

85 % (approx.)
Fraud-determination error rate (2013-2015)
34000 people (more than)
People wrongly accused of fraud (2013-2015)
20000000 USD
Class-action settlement amount (2022-2024)
78000000 USD (approx.)
Cost committed to replace MiDAS with a human-review system (2022-2024)

Details

Maturity
Discontinued
Promoter
Michigan Unemployment Insurance Agency (Michigan Department of Labor and Economic Opportunity)
Period
2013–2015 (deployment); 2015–2024 (litigation and settlement)
Keywords
unemployment insurance, fraud detection, benefits administration

Context

Michigan's Unemployment Insurance Agency deployed the Michigan Integrated Data Automated System (MiDAS) in October 2013 to automatically detect and adjudicate unemployment-benefit fraud, replacing human caseworker review with an algorithm that cross-checked claimant data against employer wage records.

Results

Investigations by Bridge Michigan, IEEE Spectrum and the state auditor general found the system issued fraud determinations with roughly an 85% error rate, automatically imposing quadrupled penalties and referring cases for wage garnishment without a human ever reviewing the underlying evidence; more than 34,000 people between 2013 and 2015 were wrongly accused, some losing homes or filing for bankruptcy.

Conclusions

A nearly nine-year class-action lawsuit (Bauserman v. Unemployment Insurance Agency) twice reached the Michigan Supreme Court; the state agreed to a $20 million settlement finalized in 2022-2024, and the Unemployment Insurance Agency committed roughly $78 million to replace MiDAS with a system that restores human review of fraud determinations.

Implementation

Indicative cost
Very high (> €5M) — The state agreed to a $20 million class-action settlement (finalized 2022-2024) and committed roughly $78 million to replace MiDAS with a system restoring human review.
Time to results
Long (> 3 years) — MiDAS deployed October 2013, active fraud-flagging 2013-2015; Bauserman v. UIA litigation ran nearly nine years, reaching the Michigan Supreme Court twice, with settlement finalized 2022-2024.
Staffing & skills
Michigan Unemployment Insurance Agency (Michigan Department of Labor and Economic Opportunity)

Conditions for success

  • Restoring mandatory human review of every fraud determination before penalties are imposed, as the state committed to do in the replacement system
  • An effective appeals process before penalties (quadrupled fines, wage garnishment) are applied
  • Independent, adversarial testing/auditing of automated fraud-detection accuracy before statewide rollout

Common failure modes

  • No human review of algorithmic fraud determinations before quadrupled penalties and wage garnishment were imposed
  • Approximately 85% of fraud determinations were later found to be erroneous
  • No effective appeals process for accused claimants before financial harm occurred
  • Nearly nine years of litigation were required before the state acknowledged and remedied the harm

Where it fits

Governance type
state government agency
Scale
state (Michigan)
Income level
high_income

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful