‘Letting the algorithm rip’: no legal basis for lack of human override of aged care funding tool, inquiry hears

There appears to be no legal barrier for a human to override a controversial algorithm that determines financial support for elderly Australians, a Senate inquiry has heard, despite government assessors being banned from doing so.The Integrated Assessment Tool (IAT), introduced in November as part of aged care Support at Home reforms, is used to assess eligibility and assign funding levels for aged care services.But Guardian Australia has reported on how the algorithm frequently under-assesses levels of need and criticisms of the IAT after the government removed the ability for assessors to override an incorrect outcome for home support.On Wednesday night, Department of Health, Disability and Ageing staff confirmed that the government has asked for advice about how to reinstate the override function, during a hearing into Support at Home.Greens senator Penny Allman-Payne asked senior Department staff about “the legislative basis for the inability to have human override as part of that process”.The Department’s first assistant secretary, Robert Day, said: “The no override comes from the fact that that is an objective outcome.”“If you have these scores from your assessment, you get this level of classification … there’s no discretionary element,” Day said.The IAT user manual includes a section on the override function, but does not clearly link any ban on its use to a specific provision of the Aged Care Act.The Liberal senator Paul Scarr has previously raised concerns about the legality of preventing assessors from overriding the outcome, telling the Senate last month that although the user manual “refers to rule 81-10 of the Aged Care Rules,” these rules “make no reference whatsoever to overriding assessments”.“These Aged Care Rules went before the delegated legislation scrutiny committee, on which I sit,” he said. “There was no suggestion that there would be these guidelines preventing overruling of assessments.”Scarr told Guardian Australia that it was “very disturbing” that a user guide could “impose automated decision-making in clearly inappropriate circumstances”.It came after a whistleblower told Guardian Australia that the algorithm frequently leaves older Australians without adequate funding and care, or downgrades their existing support.Staff could override tool during testing phaseDepartment staff also confirmed on Wednesday that testing of the algorithm was done by the Department during 2024 and 2025, but only with human override still in place. It was not tested after the ability to override was removed.The algorithm component of the IAT formed a significant part of the questioning by senators during Wednesday’s hearing.“What’s the point of having someone who’s experienced … and not giving them the ability to actually make a change if they do see an error?” the independent senator David Pocock asked.“Where is the human in all of this?”In response, Department staff said that there was no way of knowing how many times an assessor raised concerns about the outcome of the IAT, as this data was not tracked.The inquiry heard that the Department has received 834 requests for internal review of IAT outcomes since it was introduced in November. A review, which takes more than two months on average, is the only recourse if the algorithm made a mistake.“We learned last night that IAT review requests have gone through the roof since November 1, and that the legal and regulatory basis for the automation of classifications and the removal of human oversight is dubious at best,” Allman-Payne told Guardian Australia on Thursday.“It’s clear that removing human override and letting the algorithm rip is having detrimental effects on older people. When people with degenerative conditions like motor neurone disease are having their assessed care needs downgraded, something is clearly wrong.” Do you know more? melissa.davey@theguardian.com