Robodebt (Australia)

Robodebt, officially the Online Compliance Intervention, was an automated scheme that raised welfare debts in Australia. The Department of Human Services averaged a person's annual income from the Australian Taxation Office across fortnights. It compared the result with the income the person had declared, raised debts, and left the person to disprove them.

Two points in time

This network shows the same deployment as the Robodebt Scheme network, at its second documented point in time. The Robodebt Scheme network shows the scheme running. This network shows the refund work after the scheme ended, and the reform that follows it. It runs from the class-action settlement at the end of 2020 through the reform after the Royal Commission's July 2023 report.

What went wrong

The Royal Commission into the Robodebt Scheme documented hundreds of thousands of wrongful debts raised by an unlawful income-averaging method. It documented the human toll of automated debt collection against vulnerable people. It also documented the institutional failures that kept the scheme running for years despite internal and external warnings.

The Royal Commission's report, released in July 2023, found the scheme unlawful. It found that income averaging, as Robodebt used it, "was in fact inconsistent with social security legislation." It called Robodebt "a crude and cruel mechanism, neither fair nor legal."

The refunds and the settlement

After the scheme ended, about 470,000 wrongful debts were to be refunded to recipients.

At the end of 2020, the government settled a class action covering the refund of about 430,000 debts, worth more than $720 million before damages and legal fees. The Federal Court approved the settlement in 2021, in Prygodicz v Commonwealth.

Who acted

The case file records that warnings existed, but no one inside the agency was positioned, or willing, to act on them. The work of correcting debts fell on recipients, each appealing alone.

The case file says the scheme's end required the heaviest outside actors: courts and a Royal Commission. By the sources' dates, the class-action settlement came first, at the end of 2020, and the Federal Court approved it in 2021. The Royal Commission reported in July 2023.

What this case asks

This case asks what the agency needs to halt, check, and correct such a system itself. The halt, the method check, and the correction that ended Robodebt came from outside.

In this case, the reform means whoever carries the Royal Commission's recommendations into the agency. The sources read for this case do not name them. The reform can improve the engine, and add a check as it assesses each person. It can raise officers' capacity to correct debts. It can also limit how much engine output is sent to officers, and how much the engine writes to the debt register.

The Royal Commission made 57 recommendations. One was a body to monitor and audit automated decision making.

What the available tools can and cannot address

A failure pathway is a link between two parts of the network where a mistake made by one part can be passed on to the other. A tool closes a pathway when mistakes stop passing along it.

This case has a budget of 10 units and offers five tools. Upgrade model costs 3 units, or 5 at its stronger setting. Verify output costs 4. Train the staff costs 2, or 4 at its stronger setting. Pause AI on alarms costs 4. Gate record entries costs 3. These prices are the same at every target level.

Explore (No Targets) sets no targets. Under Service Targets Only, the targets are met before you apply any tool. Service here means the agency's work getting done. Two kinds of combination miss those targets. Every combination that includes Pause AI on alarms misses them, because halting the engine also halts the assessments sent to the internal review. Every combination with both Verify output and Gate record entries misses them too.

Under Service and Safety Targets and All Governance Targets, this case is not fully addressable with the available tools. Both levels ask you to close every failure pathway, among other targets. Every combination within the budget was checked, and none meets them.

Money is not what stands in the way. With every tool at its strongest setting, costing 20 units, two failure pathways stay open. They are officers reading debts for refund work, and the debts read into the proceedings.

No tool offered here acts on either of them. The first is the refund work this period exists for. The second is the courts' and the Royal Commission's reading of the debt register. That is a finding about the deployment, not a gap in your approach.