Skip to content

PAN Lab example

Robodebt (Australia)

The debt you must disprove: an income-averaging engine

Australia's Robodebt scheme raised welfare debts by spreading a person's yearly tax-office income evenly across fortnights. Recipients then had to disprove the debts.

See more

The Online Compliance Intervention, known as Robodebt, was run by Australia's Department of Human Services through Centrelink, which paid welfare benefits. It averaged yearly income from the Australian Taxation Office across fortnights and compared it with the income the department held. Where they differed, it sent a letter asking the person to disprove the debt or pay it.

How it worked

The department took each person's income for a financial year from the Australian Taxation Office. It spread that income evenly across the year's fortnights, a process the sources call averaging or smoothing. The Royal Commission found the method treated many people as though they had received income at a time when they had not.

The same method applied to every person it checked. So one flaw repeated in debts across the country, instead of showing up as scattered errors.

By Christmas 2016 the scheme was sending more than 20,000 automated debt notices a week. The measure was to be applied to more than 850,000 cases of possible overpayment, identified between 2010 and 2013.

Who had to prove what

The Royal Commission found the burden had been placed on recipients to disprove automated assessments. This reversed burden of proof was the scheme's defining feature.

Compliance officers acted on the averaged figures as debts. The case file says the correction that mattered fell to recipients appealing one by one, the people with the fewest resources.

How debts were recovered

Recovery acted on raised debts automatically. It did so before anyone independently reviewed whether a debt was ever established.

The Royal Commission's terms of reference included the use of third-party debt collectors under the scheme. The sources read here do not say whether debts were recovered by garnishment or by keeping tax refunds.

Who was warned

The case file says warnings came from inside and outside the department for years, and the scheme kept running. No one inside the department was positioned, or willing, to act on them. No one held a halt agreed in advance.

In November 2014, the Department of Social Services was advised that the proposed income averaging "did not accord with legislation". The Royal Commission's report says there is no evidence this advice was given to the Department of Human Services at the time.

Senior people were warned. In January 2015 the Human Services Minister was sent a department brief that identified a possible need for a change in the law. The Royal Commission found the department's Secretary was made aware the scheme was unlawful and "failed to act".

How it ended

The scheme's end came from outside the department, years late, through the courts.

At the end of 2020 the government settled a class action, Prygodicz v Commonwealth. The settlement covered about 430,000 debts and more than $720 million. The Federal Court approved it in 2021. Around 470,000 wrongful debts were to be repaid.

The Royal Commission into the Robodebt Scheme was set up in August 2022, after the settlement. It reported in July 2023. It found the scheme unlawful. It called Robodebt "a crude and cruel mechanism, neither fair nor legal".

Who was harmed

The harm fell on welfare recipients, including people on low incomes, students, and people with disability. The Royal Commission documented hundreds of thousands of wrongful debts. It also documented the human toll of automated debt collection against vulnerable people.

The same deployment after the Royal Commission

The Lab draws this deployment a second time, in the case titled After the Commission: repaying the debts an engine raised. There, the courts, the Royal Commission, and the repayment of wrongful debts are in place.

Where the facts come from

The facts here come from the Royal Commission's 2023 report and website, and the Federal Court's judgment in Prygodicz v Commonwealth (No 2). They also come from the Law Society Journal's article on the findings, and a Wikipedia overview of the Royal Commission.

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. The work along it may go on.

This case has a budget of 11 units. Explore (No Targets) sets no targets. Contained means the network's mistakes are corrected rather than building on each other.

Under Explore and Service Targets Only, Vet connections alone keeps the mistakes contained, for 3 of the 11 units. Under Service Targets Only it also meets the service target. That target asks for the automated system to be helping the work.

Vet connections stops mistakes passing along the automatic copy of debts from the register into recovery. Other routes exist. Escalate checks with Peer sharing rules, for 4 units, also meets the targets at both levels.

Under Service and Safety Targets the targets can be met. That level adds closing every failure pathway, and it asks that the service not be strained. The cheapest combination costs 10 of the 11 units: Understand the system, Vet connections, Escalate checks, and Gate record entries.

Understand the system costs 4 units at this level. While it is on, Vet connections costs 2 units instead of 3, and Escalate checks costs 1 instead of 2. Gate record entries costs 3. Every combination that meets these targets includes all four tools.

Under All Governance Targets the targets are not fully addressable with the available tools. That level asks for the automated system to be clearly helping the work. No combination within the budget that contains the mistakes and closes every failure pathway does that. The best leaves the system helping, but not clearly.

More money does not change that. With the budget set aside and every tool at its stronger setting, every failure pathway closes, and the automated system ends up hurting the work.

The tools aimed at the missing halt do not contain the mistakes on their own: Pause AI on alarms, Assign a challenger, and Review on schedule. None of them acts on the register or on the copy into recovery.

Stylized model of a documented deploymentPublic benefits & eligibility

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the Robodebt-class income-averaging debt engine network: 5 components and 12 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 4 assumed · 3 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

Show all 7 assumptions
  • assumed

    This example follows the pattern of automated debt-raising documented for Australia's Robodebt scheme. It does not rebuild the real system.

  • baseline

    Recipients had to disprove the debts raised against them, the Royal Commission found. So here no one inside the department pauses debt-raising.

  • baseline

    The internal review is whoever could pause the scheme. Senior people were warned. No one paused it. The example omits courts and the Royal Commission.

  • baseline

    The system applied one averaging method to every recipient. So one unlawful method raised wrongful debts across the country, not scattered errors.

  • assumed

    Recovery acts on the register's debts automatically, before any independent review. No step checks a recovery action against whether the debt was ever established.

  • assumed

    This example assumes prior debts are reused in later assessments. The sources do not describe this. They place the failure in the averaging method.

  • assumed

    Robodebt harmed welfare recipients, including people on low incomes, students, and people with disability. This example does not show them or estimate their harm.

What this example does not show

Show all 1 limitation
  • Robodebt harmed welfare recipients, and its sources record that toll. This example shows how mistakes were passed on inside the department, not the harm to people.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • The Royal Commission into the Robodebt Scheme documented hundreds of thousands of wrongful debts raised by an unlawful income-averaging method, with the onus placed on recipients to disprove automated assessments.

    empirical
    • Government Royal Commission into the Robodebt Scheme (2023) https://robodebt.royalcommission.gov.au/
    • Government Royal Commission into the Robodebt Scheme, Report (2023) https://robodebt.royalcommission.gov.au/publications/report
    • Government Prygodicz v Commonwealth of Australia (No 2) [2021] FCA 634 (Federal Court of Australia) https://robodebt.royalcommission.gov.au/publications/exhibit-2-2598-rbd999900010225-prygodicz-v-commonwealth-australia-no-2-2021-fca-634
    • Investigative Law Society Journal, Crude, cruel and unlawful: Robodebt findings https://lsj.com.au/articles/crude-cruel-and-unlawful-robodebt-royal-commission-findings/
    • Reference Royal Commission into the Robodebt Scheme (Wikipedia overview) https://en.wikipedia.org/wiki/Royal_Commission_into_the_Robodebt_Scheme
  • A single automated rule set applied uniformly and without human review produced tens of thousands of correlated wrongful fraud determinations in the documented Michigan MiDAS case — one flaw repeating at caseload scale rather than averaging out.

    empirical
    • Government Michigan AG, settlement of civil-rights class action (Bauserman, 2022) https://www.michigan.gov/ag/news/press-releases/2022/10/20/som-settlement-of-civil-rights-class-action-alleging-false-accusations-of-unemployment-fraud
    • Investigative IEEE Spectrum, Michigan's MiDAS unemployment system: Algorithm alchemy that created lead, not gold https://spectrum.ieee.org/michigans-midas-unemployment-system-algorithm-alchemy-that-created-lead-not-gold

Where this connects

Institutional pressures in this domain

  • Austerity & recovery incentives — Cost-cutting and overpayment-recovery targets tilt the system toward denial and enforcement errors.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.
  • Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.

All of them in context on the Public benefits & eligibility domain page.

Levers available here and the patterns behind them

Documented case histories