Thread 02 · Technology and Dehumanization · September 27, 2026 · 1 min read

Robodebt: when the machine accuses

In Australia, an automated calculation wrongly claimed hundreds of thousands of debts from welfare recipients. A royal commission found it cruel and unlawful.

Parliament House at dusk, Canberra ACT
Photo: Parliament House at dusk, Canberra ACT · Thennicke · CC BY-SA 4.0 · via Wikimedia Commons

The Story Granary

It was 2015-2016, in Australia.

The country entrusted an automated system with detecting welfare overpayments. It came to be nicknamed "Robodebt."

Here was the method. The system averaged the annual income reported to the tax office. Then it spread it evenly over the whole year.

As if everyone earned the same amount every fortnight.

For people with irregular incomes, the calculation manufactured debts that did not exist.

And it was then up to recipients to prove they owed nothing.

According to published estimates, hundreds of thousands of erroneous debts were issued before the system was abandoned.

Hundreds of thousands.

But recipients, lawyers, journalists and activists refused to take the machine at its word.

At the end of 2020, a class action led to a settlement. More than A$720 million was refunded. Nearly A$112 million was paid in compensation.

On July 7, 2023, the report of the royal commission chaired by Catherine Holmes was made public.

It concluded that the system had made many people feel like criminals. The commissioner summed up her findings in one sentence:

“A crude and cruel mechanism, neither fair nor legal.”
Catherine Holmes, commissioner, report of the Royal Commission into the Robodebt Scheme, 2023

The report made 57 recommendations. One of them: creating a body to oversee automated decisions.

Every person belongs to a "we," Ubuntu reminds us. When that person is reduced to an average, it is that bond we forget.

The lesson goes beyond Australia. Wherever an algorithm calculates what vulnerable people owe, there must be a human able to explain, to doubt and to correct.

A machine cannot doubt.

That duty falls to us, to each of us.

The question

When a machine decides someone owes money, who still carries the responsibility of looking that person in the eye?

Pass this story on

Figures

57recommendations from the royal commission, 2023
A$112Mpaid in compensation after the class action

Sources

  1. Prime Minister's statement on the royal commission report (Prime Minister of Australia, 2023) pmtranscripts.pmc.gov.au
  2. "Crude, cruel and unlawful: Robodebt Royal Commission findings" (Law Society Journal, 2023) lsj.com.au
  3. "Australia Robodebt royal commission hands down findings" (JURIST, 2023) www.jurist.org

Every figure on this page links to a verifiable public source.

Tags

Read also

Thread 02

A resemblance is not proof

In Detroit, a Black father was wrongly arrested in front of his daughters because of facial recognition software. His courage changed the police's rules.

Read
Thread 02

When the "nationality" box becomes a suspicion

In the Netherlands, a tax algorithm treated foreign nationality as a sign of fraud. Tens of thousands of parents were wrongly accused, and a government fell.

Read
Thread 02

Errors that do not fall at random

In 2018, two researchers showed that gender recognition systems got certain faces wrong far more often. These errors do not fall at random, and they can be corrected.

Read

The daily story

Get every story by email.

One true story a day, free and ad-free.

Or follow the RSS feed.