Calling the government’s six-year ‘Robodebt’ debt-recovery saga a “massive failure of public administration”, a Federal court ruling has saddled the government with a $1.8 billion settlement – and cemented its “shameful” data-driven debt system as a case study of the dangers when government builds policies with a blind reliance on big data.

The controversial ‘income averaging’ technique – which extrapolated ATO income data against the Centrelink benefits paid to bill 433,000 people for $1.73 billion in claimed overpayments – was declared illegal last year and its widespread use for the program has been universally panned.

Final judgement in the matter came this month as Federal Court Justice Bernard Murphy slammed the program, and the penalties in the class-action litigation it spawned, as a “shameful chapter” in the government’s welfare legacy and “a huge waste of public money”.

Although it has already repaid $751 million in invalid debts, forgiven $268 million in debts that had not yet been paid and dropped claims of $744 million in debts that had been partially repaid, the government has still refused to concede legal liability to members of the class action.

Legal action may well continue as 680 people have opted out of the settlement and could launch their own claims against the government.

Who watches the watchers?

Yet behind the eye-watering figures and tragic anecdotes from victims of the scheme lies a longer-term warning about the dangers of government leaning so hard on automated policymaking and enforcement that human judgement is removed from the equation.

“Automation can improve the consistency and efficiency of government processes,” QUT senior law lecturer Anna Huggins wrote in The Conversation, “but if there is bias or error in the computer program or data set, a flawed decision-making logic will be applied systematically, meaning large numbers of people could be affected.”

Data matching between the ATO and other government agencies has been a particular source of concern, with reports last year that the ATO would use data matching to audit childcare rebates and COVID-19 related early superannuation withdrawals and the ATO itself maintaining a growing list of ways it is using data matching across federal government programs.

Despite government guidelines and the ATO’s transparency about data matching, however, failures of automated IT systems were continuing to raise concerns about the increasing automation of government programs that can profoundly impact the people they serve.

Human oversight of the Online Compliance Intervention (OCI) program, of which Robodebt was a part, was removed in 2016 despite warnings to the government that the automated debts could be inaccurate.

As opposed to the Australian government’s increasing reliance on automated enforcement, the European Union’s general data protection regulation (GDPR) has cracked down on automated data processing, noting that data subjects “should have the right not to be subject to a decision… evaluating personal aspects relating to him or her which is based solely on automated processing and which… significantly affects him or her, such as automatic refusal of an online credit application or e-recruiting practices without any human intervention.”

Such concerns are at odds with the industry’s breathless enthusiasm about data-driven government, championed by the likes of Deloitte as a transformative force that will leverage data-driven automation into massive savings across government.

Realising those benefits, Deloitte noted in a recent analysis, requires surmounting barriers including the need for social intelligence, creative intelligence, and perception or manipulation – all of which will require ongoing “‘cognitive collaboration’ between humans and machines.”

“‘Being digital’ is not only about deploying sophisticated technology,” the firm’s analysis warns. “In addition to building sophisticated technological capabilities, ‘being digital’ requires a mindset shift – and the talent and cultural capabilities that go along with it.”

Minds the gaps

Data-driven decision making has become the ‘new normal’ post COVID but governments must be realistic about their data maturity – and must, Deloitte advises, build and sustain public trust in governments’ data-driven transformation.

Yet this can be difficult for government agencies where, US Department of Defense director of CFO Data Transformation Greg Little told a recent panel discussion about digital government, “oftentimes, we confuse our data strategy with a technology strategy.”

“Technology is an important part of realising the power of data to improve outcomes,” he said, “but technology is a tool – and there’s really an ecosystem that goes around technology to make sure we’re using it in the appropriate way to actually get the outcomes that we’re looking for.”

Those outcomes were difficult to achieve when many government agencies were still wrestling with skills challenges, the relatively immature public sector data culture, and scattered islands of legacy data that had, Howard Levenson, founder of data-analytics firm Databricks Federal added, left a persistent “gap” between the ideals of data-driven government and its reality.

“You can imagine how difficult it is to build a federal data strategy with all of these islands of data,” he explained. “You can’t use AI and machine learning until you’ve federated all of that data together, and you’ve normalised all the data before you actually start applying advanced analytics.”

“Hopefully, from those analytics we’ll be able to deliver better results that may come in the form of better national security, reduced financial crimes, better social services, and many other capabilities that the government can benefit from with better use of data.”

Labor’s Government Services spokesperson Bill Shorten expressed anger at the lack of the government’s accountability.

“No senior public servant has lost their job, no minister has lost their job,” he said.

Ministers Stuart Robert, Christian Porter, Alan Tudge, and Anne Ruston all at some stage were responsible for the social and government services portfolios during Robodebt’s operation.