The Murphy Campaign is putting a new question to the Australian Government:
Who actually decides how much care an older Australian needs — the human assessor, or the algorithm?
Australia's aged-care assessment system uses the Integrated Assessment Tool and algorithm-supported classification to help determine Support at Home classifications.
Technology itself is not the enemy. Used properly, technology can improve consistency, speed and access.
But when an automated decision-making system materially influences the care available to an older Australian, secrecy, weak explainability and restricted human intervention become matters of public accountability.
Recent public reporting and expert analysis have raised disturbing questions about the system.
The Integrated Assessment Tool collects hundreds of pieces of information about an older person. Yet expert analysis reported by ABC Four Corners found that much of the information collected does not feed into the classification algorithm.
Assessors can also record extensive free-text information describing the human being sitting in front of them — their circumstances, risks and needs — yet that material is not necessarily read by the algorithm producing the classification.
That leads to an extraordinary question:
Why collect information about a human being if it cannot materially influence the machine assessing their level of support?
A single point can mean $40,000 a year
Professor Kathy Eagar's analysis reported that factors including cognition, frailty and mental health can enter the decision tree relatively late. At one point in that decision process, a difference of a single point can reportedly correspond with a difference approaching $40,000 a year in available funding.
Think about what that means.
Two older Australians can have complex and broadly similar lives, yet a seemingly small difference at a decision boundary may contribute to dramatically different funding outcomes.
And historically, an assessor who knows the person, sees their circumstances and believes the resulting classification does not adequately reflect their needs has faced constraints on simply overriding that outcome on the spot.
That is where technology stops being merely an administrative tool and becomes a question of power and accountability.
A machine-made shortfall becomes a "Heat or Eat" choice
A funding gap on a spreadsheet is an abstraction. A funding gap in an 85-year-old's fortnightly budget is not.
When an opaque, algorithm-driven classification under-funds an older Australian's actual level of need, that shortfall does not disappear. It is absorbed — by the person, by their family, or by their health.
It is absorbed by a pensioner turning off the heater on a cold night because home care hours were classified too low to cover both warmth and groceries. It is absorbed by a family member paying out of pocket for support the system decided wasn't needed. It is absorbed by an older Australian skipping a meal, or skipping medication, to make the fortnight stretch.
This is the "Heat or Eat" travesty — and an unaccountable algorithm is helping to manufacture it.
Older Australians on fixed incomes already ration heating, food and medical care against the cost of living. A classification system that quietly under-counts a person's real needs — because a free-text clinical note wasn't read by the algorithm, or because cognition and frailty entered the decision tree too late — does not just produce a wrong number. It pushes a real person closer to that impossible choice between heating their home and feeding themselves properly.
No government should be able to say a classification was "validated" while the human consequence of getting it wrong is an older Australian going without heat, without food, or without care. That is not an edge case. It is the predictable result of a black-box system making high-stakes decisions about people who are already living on the margin.
The machine doesn't have to be evil to be dangerous
The danger is not a computer sitting somewhere maliciously deciding to hurt older Australians.
The danger is much more mundane — and potentially much more powerful.
Responsibility can disappear into the system.
The assessor can say: "The system produced the classification."
The Department can say: "The algorithm was validated."
Government can say: "You can request a review."
And the older Australian is left asking:
WHO ACTUALLY DECIDED HOW MUCH CARE I NEED?
That question must have an answer. A government should never be able to hide public responsibility behind technological complexity.
You cannot meaningfully challenge what you cannot understand
Review rights matter. But a person cannot effectively challenge a decision unless they can understand the material reasons the decision was made.
Which answers mattered? Which answers didn't? What threshold was crossed?
What evidence was disregarded by the automated process? Could the assessor intervene?
What version of the algorithm was used? Has that algorithm subsequently changed? How was it independently validated? How often are its classifications changed following human review?
These are not unreasonable demands. They are elementary requirements of accountable government.
SHOW US THE RULES.
The Murphy Campaign is calling for genuine sunlight over automated decision-making in aged care. We demand:
Show us the rules
Publish the classification methodology, relevant decision logic, variables and thresholds to the fullest extent lawfully possible.
Show us what counts — and what doesn't
Tell older Australians which information collected during their assessment materially influences classification and which information does not.
Show us the human override
There must be meaningful clinical escalation and human intervention when an algorithmic result fails to reflect the person standing in front of the assessor.
Show us the explanation
Every person affected should receive an intelligible explanation of the material factors that produced their classification.
Show us the version history
When the algorithm changes, disclose what changed, when it changed, why it changed and what validation occurred.
Show us the review data
Publish how frequently classifications are reviewed, changed or overturned and why.
Show us the validation
Permit genuine independent examination of whether the classification system is clinically sound, fair, explainable and sufficiently responsive to cognition, frailty, mental health, carer circumstances and other complex needs.
This is bigger than Sheila
The Murphy Campaign began with the lived experience of Sheila Murphy and one family's struggle through Australia's aged-care system. But Sheila's experience forces a much larger question.
Australia is building increasingly sophisticated systems to assess, classify and allocate public services. That may be inevitable. Unaccountable automated government is not.
If an algorithm influences whether an older Australian receives thousands of dollars more or less in support, Australians have a right to understand the basis of that outcome. If clinical information is collected but does not influence the classification, Australians deserve to know. If a human assessor cannot correct an outcome they believe is inappropriate, Australians deserve to know.
And if government believes its algorithm is fair, clinically valid and delivering better outcomes, then transparency should strengthen public confidence — not threaten it.
Put the algorithm in the sunlight.
The Senate Community Affairs References Committee is examining the Support at Home program, with its next public hearing scheduled for 22 September 2026 in Canberra. This is precisely the kind of question parliamentary scrutiny should examine.
The Murphy Campaign calls on Parliament, the Government, oversight agencies, aged-care advocates, clinicians, unions, researchers, journalists and older Australians themselves to demand answers.
Not rumours. Not conspiracy theories.
Evidence.
Because when government technology influences a person's care, the burden should never fall on an 85-year-old Australian to decipher the machine. The burden belongs to government to explain it.
THE ALGORITHM MUST ANSWER TO THE PERSON.
SHOW US THE RULES. · SHOW US WHAT COUNTS. · SHOW US THE HUMAN OVERRIDE.
SHOW US THE EXPLANATION. · SHOW US THE VERSION HISTORY. · SHOW US THE REVIEW DATA. · SHOW US THE VALIDATION.
Sunlight. Evidence. Accountability. Action.
— Murphy Campaign
Approved for Care. Waiting for Care.
