The Pulse
Australia’s Ombudsman Finds Aged-Care Algorithm Unfair
Australia’s Commonwealth Ombudsman says the automated tool used to assess aged-care eligibility, funding and priority lacks safeguards and undermines confidence in fair decisions. Iain Anderson recommends changing the law to let assessment
AI.info Team · Written with AI · fact-checked ·
“A reasonable person would, in my view, lack confidence that the system is in fact fair and consistent.”
Iain Anderson, Commonwealth Ombudsman
A decision with no human override
In a report released on October 26, Commonwealth Ombudsman Iain Anderson found that the Department of Health, Disability and Ageing acted unreasonably in administering Australia’s Single Assessment System, including the automated tool used in home-care assessments. The report recommends that the government amend the law so assessment delegates can replace an automated result in appropriate circumstances. Under the current framework, the system’s automated decisions cannot be changed by a human, even when considered clinically inappropriate.
The Integrated Assessment Tool uses answers and observations gathered by an assessor to produce decisions about eligibility, a person’s ongoing funding level and, for Support at Home, their priority category. Assessors then use those outcomes to guide recommendations for services. Anderson says that leaves a mismatch in the process: human judgment shapes the information entered, but a human cannot override the resulting decision.
The report describes a complaint involving a person with a rapidly progressive, incurable neurological condition. The person had been approved for the second-highest level of home-care support but was given a standard wait time of 10 to 11 months; the complaint said the algorithmic process blocked an escalation despite the department acknowledging the outcome was inappropriate in the circumstances.
The department’s testing and oversight drew criticism
Anderson found that the department introduced the tool in its current form without meaningful risk assessment, evaluation or analysis. The report says public explanations had not meaningfully shown how the algorithm turned assessment inputs into outcomes, and that the department had not met transparency recommendations made by the Royal Commission into the Robodebt Scheme.
The department disputed the criticism that it had not tested or validated the system. As ABC News reported, testing had been conducted on an earlier algorithm developed by a consultancy; the department later modified the version now in use, which had no external validation. Anderson also faulted oversight of the contracted assessment providers, citing gaps in checks on assessors’ training and qualifications, weak use of complaints to identify systemic problems, and provider performance falling short in important areas.
The Ombudsman did not assess whether the algorithm’s outputs were clinically appropriate, and said the investigation did not establish that it produces clinically inappropriate results. Instead, the report says its findings raise serious questions about reliability and suitability and warrant independent clinical review. That distinction matters: the report condemns the system’s design, transparency and administration without claiming to have independently validated each funding outcome.
Six recommendations accepted; human override only noted
Anderson made 10 recommendations, including a legal route for human substitution, a review of the Act and rules, clearer public explanations, stronger audits of assessment providers and a better complaints framework. ABC News reported that the government accepted six recommendations, while noting two—including the proposed human override and related changes to the aged-care rules. The department accepted a review of the legal framework and public consultation on future amendments.
The report also records steps already taken: the government announced an escalated review pathway for certain assessment decisions on October 16, and the department published information about the classification algorithm and its business rules on October 6. Anderson welcomed those actions but said it was too early to know whether the published material lets independent parties reproduce assessment outcomes.
On October 27, MND Australia called for the government to suspend use of the tool to set Support at Home funding, arguing that people with motor neurone disease cannot wait for further reviews. That is an advocacy group’s demand, not a government decision. The report’s central remedy remains legislative: until the law changes, assessment delegates cannot substitute their judgment for an automated result.