If you're going to build the biggest algorithmic debt-chasing machine in the country, you'd better do it right.
If not, you risk raising the ire of the nation's most vulnerable people -- and data scientists.
Centrelink has been under fire for months for sending out automated debt notices, based on data scraped from the Australian Tax Office and Centrelink's own payment records, spawning the #RoboDebt hashtag.
The government hopes the new system will help recover AU$4.5 million in welfare debt every day. But critics say the system is a blunt instrument that is incorrectly sending out automated notices, often to people who don't owe any money, with little human oversight.
One of those critics is Daniel Angus. After receiving a routine debt notice from Centrelink over his Family Tax Benefit payments, the data scientist and academic says he "hit the roof."
And that's when on Twitter -- as they say in computer science terms -- shit got real.
Centrelink replied to Angus, notifying him that "the Employment Income Matching measure does not apply to Family Tax Benefit Payments."
But after seeing the #Robodebt issue play out in the news over recent months, Angus decided to start asking questions. Complex questions. Data scientist questions.
Angus says he was being tongue-in-cheek, but speaking over the phone from his office at the University of Queensland, it's clear he's annoyed.
"[We] expect answers on questions of human process, why can't we expect answers for those digital processes? Those questions are valid questions to be asked of an algorithmic process."
While he's still working out the details of his notice, he says the Robodebt debacle has been concerning since day one.
"I think it is an absolute prime example of the worst of analytics," he says.
"This is the kind of thing you see when you have someone behind the scenes who's treating this just as data, and not considering that everyone point within this system is a person... they're not just rows in a spreadsheet, there are people behind this."
The Department of Human Services, which is responsible for Centrelink, said it did not have further comment to provide, other than the tweets it sent to Mr Angus. The Department did not respond to specific questions from CNET about the setup or oversight of its data matching systems and the #Robodebt fallout.
Instead, the Department pointed to its submission to the Senate Inquiry on Centrelink debt collection, which says new income matching measures have not changed Centrelink's data-matching methodology or the way it calculates welfare debts.
Additional reporting by Luke Lancaster.
First published April 7, 3:15 p.m. AEST.
Update, April 10 at 9:41 a.m.: Adds comment from Centrelink.
Tech Enabled: CNET chronicles tech's role in providing new kinds of accessibility.
Technically Literate: Original works of short fiction with unique perspectives on tech, exclusively on CNET.