By Andrew Paul Klein

Published – 1 October 2026

Introduction

This is the first chapter in a series examining how the Australian state replaced human judgment with algorithmic governance in the delivery of human services.

The series is not an argument about technology. It is an argument about ideology. The instruments — the assessments, the classifications, the scoring systems, the automated determinations — did not arrive on their own. They were preceded by a decision about how the state should relate to its citizens. That decision was made decades ago, and it was made in the name of efficiency.

The decision was to treat human services as a market.

Once that decision was taken, everything else followed. Markets require measurement. Measurement requires instruments. Instruments require standardisation. And standardisation requires that the person be translated into data.

This chapter traces that sequence. It begins with the ideological shift that began in the 1980s and accelerated under the Howard and Kennett governments. It examines the institutional architecture that translated the ideology into policy — the Productivity Commission, the market-based schemes it designed, and the restructuring of the public service that accompanied them. It then explains the structural logic that connects the market model to the instruments that would come to define how the state sees its citizens.

The argument is not that the people who designed these reforms were malicious. It is that they were ideological — in the precise sense that they operated within a framework of assumptions they did not examine. The framework treated markets as neutral mechanisms for allocating resources. It treated competition as a driver of quality. It treated measurement as a form of accountability. Each of these assumptions was plausible. None of them was contested.

The consequence is the subject of this series. When the state becomes a market designer rather than a service provider, it must measure what the market produces. When it measures, it must standardise. When it standardises, it must translate human need into data. And when human need becomes data, the human being is no longer in the system.

They are in the gap between what the instrument measures and what they actually are.

That gap is where the harm occurs. And the gap was not an accident. It was the logical consequence of a decision about how the state should relate to the people it serves.

This chapter documents that decision. The chapters that follow document its consequences.

What This Chapter Covers

The ideological shift. From the Hawke-Keating reforms through the Howard and Kennett governments, Australia embraced a set of policies that prioritised market mechanisms over universal provision. The shift was not a single event. It was a series of decisions, each defended on its own terms, that accumulated into a new default.

The institutional architecture. The Productivity Commission became the central institution for designing market-based schemes in human services. It played a central role in the design of the NDIS, the Child Care Subsidy, and the employment services market. Its terms of reference were explicit: identify which services were best suited to competition, contestability, and user choice.

The restructuring of the public service. The Howard government used the creation of Centrelink to restructure the APS around commercialised service delivery. The enabling legislation described the agency’s functions in terms of “service arrangements” and “directions from the minister” — not in terms of a duty to citizens. The culture of “frank and fearless” advice was eroded.

Why measurement became necessary. Markets require comparison. Comparison requires standardisation. Standardisation requires instruments. The instruments — the Job Seeker Classification Instrument, the Star Ratings system, the NDIS assessment tools — became the lens through which the state perceives its citizens.

The pre-filtered information problem. Decision-makers do not see the raw data. They see the output of the instruments. The documentation of harm — the evaluations, the reviews, the audits — is externally generated. It does not enter the decision-making architecture unless it becomes a political liability.

The chapter concludes with the structural insight that organises the rest of the series: the state has not been captured by individuals. It has been reconfigured around a method. The method was chosen by an ideology. The ideology was implemented through markets. The markets required measurement. The measurement became the lens. And the lens does not see the person.

A Note on Method

This series follows the same discipline as the work that preceded it. Every claim is sourced. Where a claim rests on inference rather than direct evidence, it is marked as such. Where a source cannot be verified, it is marked as unverified. The reader is encouraged to check the references, test the argument, and reach their own conclusion — including the conclusion that the author is wrong.

The series is not a polemic. It is a structural analysis. The distinction matters, because polemic invites agreement or disagreement, while structural analysis invites investigation. The question is not whether the reader shares the author’s politics. The question is whether the architecture described in these chapters is real, and whether its consequences can be traced in the record.

They can. The record is extensive. It includes Royal Commissions, parliamentary inquiries, audit reports, academic studies, and the testimony of the people who lived through the systems. This series draws on that record. It does not invent. It documents.

The Chapters

Chapter 1: The Neoliberal Turn — How Market Ideology Entered Human Services and Why It Required Measurement

Chapter 2: The Instruments — How the State Began to See Citizens as Data Points

Chapter 3: Robodebt — The Paradigm Case

Chapter 4: The NDIS — Epistemic Injustice and the Datafication of Disability

Chapter 5: Aged Care — The Integrated Assessment Tool and the Outsourcing of Care

Chapter 6: Veterans — The Royal Commission and the System “Geared to Frustrate”

Chapter 7: The NSW Mental Health System — Financial Management Orders and the Pipeline to the State Trustee

Chapter 8: The Proxy Model Domesticated — How the State Uses Providers to Mediate Its Relationship with Citizens

Chapter 9: The Lens That Cannot See — What the Algorithmic Gaze Shows and What It Hides

Chapter 10: The Question — Can the State Be Redesigned to See, or Has the Lens Become the State?

Chapter 1: The Neoliberal Turn — How Market Ideology Entered Human Services and Why It Required Measurement

Method notes. Claims are classified throughout as Established, Inference, or Speculation.

1.1 The Ideological Shift That Preceded the Machines

The algorithmic lens did not arrive in a vacuum. It was preceded by an ideological transformation that made it not merely possible, but necessary.

From the 1980s onward, Australian governments of both major parties embraced neoliberalism — a set of policies that prioritised market mechanisms, competition, and efficiency over universal provision. The Hawke-Keating Labor governments opened the economy, deregulated finance, and embraced competition policy. John Howard’s Coalition government (1996–2007) took that architecture as given and translated it into something more culturally fluent and politically durable: aspiration, reward, “choice,” and the quiet moralisation of the market as common sense.

As one analysis puts it, Howard “did not simply reject Labor’s reform era; he normalised it – then tilted it further, especially where labour power and public provision might have tempered inequality”. The deeper legacy was not a bundle of policies but a set of defaults — assumptions both major parties now treat as common sense.

At the state level, Jeff Kennett’s Victorian government (1992–1999) was the pioneer. The shift to “entrepreneurial government and the market state” reflected the dominant paradigm of economic rationalism, illustrating Australia’s interconnectedness with similar reforms in New Zealand, Canada, the USA, and the UK. The Kennett government’s approach was explicit: privatisation originated as a political and financial strategy, and “the economic rationale was appended later”.

1.2 The Productivity Commission as Market Architect

If neoliberalism was the ideology, the Productivity Commission was its institutional architect.

The Commission has “led analysis on why, when and how market-based competition can improve human services, playing a central role in the design of key market schemes, including the NDIS and Child Care Subsidy”.

In 2016, the Commission was formally tasked by the Turnbull government with examining “the application of competition and user choice to services within the human services sector” and developing “policy options to improve outcomes”. The terms of reference were explicit: the Commission was to identify which services were “best suited to the introduction of greater competition, contestability and user choice”.

The Commission later attributed “poor incentives and poor outcomes” to “weak market design and insufficient government oversight,” and recommended that government should “recognise and actively engage in its role as a market designer and market steward”.

This is the structural transformation in institutional form: the state is no longer a provider of services. It is a market designer.

1.3 The Howard Government’s Restructuring of the Public Service

The ideological shift was accompanied by a structural transformation of the Australian Public Service itself.

The Howard government used the creation of Centrelink to “restructure the Public Service to commercialise, corporatise or privatise the welfare service delivery function”. The enabling legislation described the new agency’s functions in terms of “service arrangements” and “directions from the minister” — not in terms of a duty to citizens.

Kathy MacDermott’s book Whatever Happened to Frank and Fearless? (2008) plots “the changes in the culture of the Australian Public Service that have led many contemporary commentators to lament the purported loss of traditional public service values of impartiality, intellectual rigour and – most importantly – the willingness of public servants at all levels to offer frank and fearless advice to their superiors and their ministers”.

MacDermott’s central argument is precise: “structural and cultural change compromising the integrity of the public service reached its apogee towards the end of the eleven years of the Howard government”.

The consequences are documented. A submission to the Australian Public Service Review notes that “since 2013, there have been over 14,000 APS jobs cut” — with the consequence of driving up consultancy expenditure, creating a capability gap, and making the APS “dependent on these consultants”.

The submission identifies a cultural transformation: “making budget savings is held in higher regard than quality of service delivery,” with “an expectation of punitive response from Government for attempting provision of service excellence by reinvestment of budgetary savings”.

The result is a public service that can “police and procure but struggles to build”.

1.4 Why Market Models Require Measurement

Here is the structural connection between neoliberalism and the algorithmic lens.

To introduce competition, you must be able to compare providers. To compare providers, you must standardise outcomes. To standardise outcomes, you must quantify need.

The Productivity Commission’s approach to human services reform was explicitly about creating markets where “informed user choice” could operate. But choice requires information. Information requires measurement. And measurement requires instruments.

This is not a criticism of the Commission’s intentions. It is a statement of structural logic. If the state is to function as a market designer rather than a provider, it must be able to see the market — and what it sees must be quantifiable.

The instruments followed:

· Job Seeker Classification Instrument (JSCI) — a questionnaire of up to 49 questions that determines a job seeker’s “level of disadvantage” and places them into one of three streams.

· Star Ratings — a performance management system that gives providers a rating “based on their comparative performance in achieving employment or educational outcomes”.

· NDIS budget-setting tools — independent assessments that reduce disability to measurable criteria.

Each of these instruments is a lens. Each of them shapes what the state can see.

1.5 The Job Network as Laboratory

The Howard government’s Job Network was the first large-scale experiment in applying market principles to human services.

The system operated as a “black-box market“: jobseekers were referred by Centrelink to Job Network providers, who “had significant discretion and were contracted for results”. Services were “dependent on assessed needs through the Job Seeker Classification Instrument (JSCI) score”.

In 1999, the government introduced “biannual Star Ratings system for performance evaluation — used to reward higher-performing providers and remove business from poorer performers”.

The architecture is precise:

1. The instrument (JSCI) measures need.

2. The market (Job Network) allocates resources based on the measurement

3. The performance system (Star Ratings) rewards providers based on outcomes.

4. The state sees the data, not the people.

This is not a system that failed to see the human. It is a system that replaced the human with the instrument.

1.6 The Pre-Filtered Information Problem

The consequence of this architecture is structural, not accidental.

When a minister looks at employment services, they see Star Ratings, outcome figures, and cost-per-outcome estimates. They do not see the jobseeker who was streamed incorrectly, the provider who gamed the system, or the person who fell through the gap between the instrument’s categories.

The documentation of failure — the evaluations, the reviews, the audits — is real. But it is externally generated. It comes from outside the decision-making architecture.

The state processes the data that the instruments generate. The instruments generate the data that the market design requires. And the market design was chosen by the ideology.

1.7 The Structural Capture

The capture is not of individuals. It is of architecture.

The Productivity Commission designs the markets. The markets require instruments. The instruments produce data. The data feeds the decision-makers. The decision-makers are accountable to the market’s logic, not to the citizen’s experience.

This is what we mean by structural capture. The state has not been bribed. It has been reconfigured.

The reconfiguration was ideological. It was implemented through institutional design. And it required the technology that would make it operable.

The human lens — caseworkers, assessors, reviewers — was not removed by malice. It was displaced by an architecture that did not need it.

Claim -Status -Summary

#- Claim- Status

1 -Hawke-Keating opened the economy; Howard normalised and extended neoliberalism -Established 

2 -Kennett pioneered state-level market reforms in Victoria- Established 

3 -Productivity Commission designed NDIS and Child Care Subsidy market schemes -Established 

4- Howard restructured public service to commercialise welfare delivery- Established 

5- APS lost “frank and fearless” culture under Howard -Established 

6- 14,000 APS jobs cut since 2013; reliance on consultants increased- Established 

7 -Job Network used JSCI and Star Ratings as instruments -Established 

8- Market models require measurement, which requires instruments -Inference

9 -The instruments became the lens through which the state sees -Inference

10- Structural capture of architecture, not individuals- Inference

Chapter 2: The Instruments — How the State Began to See Citizens as Data Points

2.1 The Shift from Judgment to Measurement

Chapter 1 established that neoliberalism required measurement. This chapter examines the instruments that perform the measurement — and the consequences of their operation.

The instruments are not neutral. They are the lens through which the state perceives its citizens. What the lens shows, the state can see. What the lens does not show, the state is structurally incapable of seeing.

This is not a metaphor. It is a description of how automated decision-making systems function. They encode specific assumptions about what matters, what can be measured, and what should be done. The assumptions are embedded in the code. The code produces the data. The data feeds the decision-makers.

The human being — with all their complexity, context, and contradiction — is not in the data. They are in the gap between what the instrument measures and what the person actually is.

2.2 The Job Seeker Classification Instrument (JSCI)

2.2.1 What the JSCI Is

The Job Seeker Classification Instrument (JSCI) — also called the “Job Seeker Snapshot” — is a set of questions a job seeker is asked in an interview with the Department of Human Services (DHS) or their employment services provider.

The JSCI “looks at a job seeker’s likelihood of becoming or remaining long term unemployed”. It has up to 49 questions. Generally, “a job seeker who has a higher level of disadvantage will answer more questions. The minimum number of questions a job seeker will answer is 18”.

2.2.2 What the JSCI Measures

Eighteen factors are assessed:

· Age and gender

· Geographic location

· Recency of work experience

· Proximity to a labour market

· Job seeker history

· Access to transport

· Educational attainment

· Phone contactability

· Vocational qualifications

· Disability/medical conditions

· English proficiency

· Stability of residence

· Country of birth

· Living circumstances

· Indigenous status

· Criminal convictions

· Indigenous location

· Personal factors

Each factor “is assigned a numerical weight, or ‘points’, indicative of the factors that make it difficult to place a job seeker into employment; their associated points are added together to calculate the JSCI score”.

“The higher the JSCI score, the higher the likelihood of the job seeker remaining unemployed for at least 12 months”.

2.2.3 The Structural Consequence

The JSCI transforms a human being into a score. The score determines the level of servicing the job seeker receives. It determines which “stream” they enter. It determines what the provider is paid.

What the JSCI cannot measure:

· The job seeker’s motivation, resilience, or determination

· The specific circumstances of their local labour market beyond statistical averages

· The quality of the relationship they might form with a caseworker

· The non-linear ways in which people find work

What the JSCI does measure:

· A set of factors that correlate with long-term unemployment

· A numerical weight assigned to each factor

· A total score that determines resource allocation

The instrument does not see the person. It sees the profile. And the profile is what the state uses.

2.3 The Star Ratings System

2.3.1 What Star Ratings Are

The Star Ratings system was introduced by the Howard government in 1999 to evaluate the performance of Job Network providers. The system gives providers a rating “based on their comparative performance in achieving employment or educational outcomes”.

2.3.2 How Star Ratings Work

The methodology is complex. Access Economics’ review of the Star Ratings method describes the process:

1. Calculate the ratio of observed positive outcomes to predicted positive outcomes for each provider in each region

2. Normalise the distributions of ratios for each Key Performance Indicator (KPI)

3. Calculate a weighted sum across KPIs using predetermined weights

4. Rank providers and apply star ratings based on a predetermined distribution

The model adjusts for “labour market conditions and job seeker characteristics” using regression analysis. The intention is to compare providers on a “like-with-like” basis, so that a provider with a caseload of disadvantaged job seekers is not penalised relative to a provider with easier-to-place clients.

2.3.3 The Problems the Model Creates

The literature on performance-based contracting identifies two structural problems: creaming and parking.

Creaming refers to “focusing on participants most likely to generate an outcome.” Parking refers to “providing little assistance to clients with low employment probabilities”. In a performance-based contract, providers are paid by results. Outcome-based indicators “encourage providers to maximise profits by getting the best outcomes as measured by the contract indicators”.

The Star Ratings model was designed to mitigate creaming and parking by adjusting for participant characteristics. But stakeholders identified other problems: “Providers do not have certainty over expected performance levels… lack of transparency over the calculations used to assess provider performance… The Star Ratings adjusts for differences in external factors that affect the provider performance using regression analysis. This methodologically demanding process may be difficult to understand”.

The instrument is opaque. Providers are ranked, rated, and reallocated business based on a calculation they cannot fully understand. The human beings they serve are even further from the calculation.

2.4 The Integrated Assessment Tool (IAT) for Aged Care

2.4.1 What the IAT Is

The Integrated Assessment Tool is “a structured digital assessment used during a home interview with an older person seeking government-subsidised care”.

Assessors “enter information about mobility, cognition, daily living and the person’s broader circumstances”. The tool “converts these inputs into scores. It then applies rules to sort the person into one of eight funding classifications”.

“Assessors are barred from overriding the tool’s classification except in a small set of pre-defined circumstances”.

2.4.2 The Opacity Problem

“Despite requests for the technical specifications, and for the identity of the team that designed the classification logic, details have not been released”.

The health department insists the tool “is not artificial intelligence (AI). It says it is a rules-based classification algorithm, not a machine-learning model”. But as the analysis notes: “whether a system involves AI or not is beside the point. Deeper issues sit underneath the technology”.

2.4.3 The Ethical Problems

The key ethical problems in algorithmic decision-making relate to opacity, discretion and accountability:

· Opacity: Can the people affected see how the system works?

· Discretion: Can the professionals using it exercise judgement?

· Accountability: Can anyone be held to answer for its decisions?

The IAT addresses none of these. The classification logic is hidden. The assessor’s discretion is stripped away. The accountability is diffused across the system.

As the analysis puts it: “A standardised tool does deliver consistency: everyone is processed the same way. But consistency is not fairness, especially when the standard is hidden and applied to people whose needs do not fit standard categories”.

2.4.4 The Loss of Street-Level Bureaucracy

The analysis cites Michael Lipsky’s classic 1980 study Street-Level Bureaucracy: “discretion is a defining feature in frontline public service work. Teachers, social workers, nurses and aged-care assessors exercise judgement precisely because rules are always incomplete, resources are constrained, and every client is unique. Strip discretion out of these encounters, and the assessment can no longer respond to what it finds, only to what the tool allows”.

Aged-care assessors describe feeling “handcuffed” by the new system.

2.5 The NDIS Assessment Reforms

2.5.1 The Shift to Automated Planning

From mid-2026, a new assessment tool will be introduced for NDIS participants. “The data from which will be fed into software to generate a budget. The NDIA delegate’s role will be limited to that of accepting or rejecting the budget”.

Dr Georgia van Toorn, writing for the UNSW Disability Innovation Institute, describes the consequences:

“There will be no safety net – no scope for either the NDIA delegate or the administrative appeals tribunal to amend the budget. In other words, the computer will have the final say”.

2.5.2 The Uneven Distribution of Harm

“The people who are most likely to have their funding cut or supports misjudged are the people whose disabilities are complex, fluctuating or not easily captured through standardised assessments”.

“A more automated process may be particularly perilous for people whose disabilities are affected or compounded by social factors – such as poverty, racial, sexual or gender discrimination, and entanglements with the carceral system – which fall well beyond the bounds of data captured by existing assessment tools”.

“Just like with robodebt, the system will inflict its heaviest toll on the communities least able to bear it”.

2.6 The Robodebt Instrument

2.6.1 The Mechanism

The Robodebt system was “the policy utilising employment income matching, a process of data-matching identified through a comparison of ATO and Department of Human Services (DHS) data”.

The system operated by “DHS obtaining information from the ATO as to what income had been declared by an income support payment recipient’s employer in the relevant financial year and comparing this information, through an automated system, with the income declared to DHS by the recipient”.

“Any discrepancy was then required to be explained by the recipient through an online system. If the recipient did not explain this discrepancy, or did not respond, or they agreed with the data, the amount declared to the ATO in the relevant financial year would be averaged across a period of time, usually the period of employment indicated by the employer, to raise a debt”.

2.6.2 The Instrument’s Assumptions

The Robodebt instrument assumed:

· That annual income could be averaged to produce a fortnightly figure

· That the averaged figure was an accurate representation of actual earnings

· That the burden of proof lay with the recipient to disprove the debt

Each assumption was wrong. Income averaging “obscured the fluctuations in actual fortnightly earnings that determine welfare entitlements”. The reverse onus placed the burden on people who often lacked the resources to challenge the system.

The result: 866,857 cases of possible overpayment, and 470,000 unlawfully raised debts amounting to $1.751 billion.

2.7 The Scale of the Lens

The instruments are not confined to welfare, aged care, and disability. The NSW Ombudsman’s Compendium of ADM Systems documents the breadth of deployment across the NSW public sector.

Examples include:

· NSW Air Quality Alert System — a “fully automated, rule-based system” that issues alerts when air quality exceeds standards 

· NSW Air Quality Forecasting System — forecasts air quality up to 72 hours ahead 

· LARRY — checks a customer’s eligibility for energy rebates using Centrelink eServices 

· FER — checks eligibility for the Family Energy Rebate 

· Energy Bill Relief — automatically determines eligibility for the National Energy Bill Relief payment 

The NSW Ombudsman’s research found 136 ADM systems reported by NSW government departments and agencies, with a potential increase of 50 per cent in the next three years.

A majority of state government departments (46 of 77) reported using or planning to use ADM systems.

2.8 The Transparency Deficit

The OAIC’s review of 23 Commonwealth agencies authorised to use ADM found:

· 4 agencies (17%) disclosed the use of ADM in decision-making in their Information Publication Scheme

· 2 agencies (9%) were “likely to be using ADM” via external sources but had not disclosed use

· 17 agencies (74%) — information to confirm whether or not ADM was in use was not able to be identified using external sources or IPS information 

74 per cent of agencies authorised to use ADM have not disclosed whether they are using it.

The NSW Ombudsman’s research identified a related problem: “Government agencies are not at this time subject to any express obligation to report on their use of ADM in the performance of statutory functions, either publicly… or when providing reasons to individuals affected by decisions that have been made by or with the assistance of ADM tools”.

The instrument is invisible. The citizen cannot see what is being done to them.

2.9 What the Instruments Cannot See

The instruments measure what they were designed to measure. They cannot measure what lies outside their categories.

The JSCI cannot measure the job seeker’s resilience, the quality of their local networks, or the non-linear path to employment. It measures 18 factors and produces a score.

Star Ratings cannot measure the quality of the relationship between provider and participant. It measures outcomes against a regression model.

The Integrated Assessment Tool cannot measure the older person’s fear, their dignity, their preference for familiarity over efficiency. It measures mobility, cognition, and daily living, and produces a classification.

The NDIS assessment tool cannot measure the fluctuating nature of complex disability, or the ways in which poverty and discrimination compound impairment. It measures against standardised categories.

Robodebt could not measure the difference between averaged annual income and actual fortnightly earnings. It measured the discrepancy and raised a debt.

The lens shows what it was built to show. The human being is in the gap.

Claim -Status- Summary

#- Claim- Status

1 -JSCI is a 49-question instrument that produces a score -Established 

2- JSCI score determines level of employment servicing- Established 

3 -Star Ratings use regression analysis to adjust for provider caseload -Established 

4 -Star Ratings model is opaque to providers- Established 

5 -Integrated Assessment Tool bars assessors from overriding classifications -Established 

6- IAT classification logic has not been released -Established 

7- NDIS automated planning will limit NDIA delegate to accept/reject- Established 

8- Robodebt used income averaging and reverse onus -Established 

9- Robodebt produced 470,000 unlawful debts totalling $1.751 billion -Established 

10- NSW has 136 ADM systems, potentially increasing by 50%- Established 

11- 74% of Commonwealth agencies authorised to use ADM have not disclosed use -Established 

12- The instruments cannot see what they were not designed to measure- Inference

Chapter 3: Robodebt — The Paradigm Case

3.1 The Scheme That Defined the Era

Robodebt is not merely an example of the algorithmic lens. It is the paradigm case. Every structural feature we identified in Chapter 2 — measurement, categorisation, opacity, the transfer of burden to the citizen — converged in a single program that operated on a national scale for four years, produced approximately 500,000 unlawful debts, and caused documented psychological harm including suicide.

The Royal Commission into the Robodebt Scheme, led by Commissioner Catherine Holmes, delivered its report in July 2023. Its central finding is unambiguous: “Robodebt was a crude and cruel mechanism, neither fair nor legal, and it made many people feel like criminals. In essence, people were traumatised on the off-chance they might owe money. It was a costly failure of public administration, in both human and economic terms”.

This chapter traces the scheme’s origin, its operation, the warnings that were ignored, the human cost, and the accountability outcome. The pattern is not one of isolated failure. It is one of structural inevitability.

3.2 The Origin: A Budget Measure in Search of a Legal Basis

3.2.1 The Political Context

In 2015, the Department of Human Services (DHS) proposed a new compliance measure as part of the federal budget. The proposal was put forward by the Minister for Social Services and began in pilot form that year before being expanded in subsequent budgets.

The political framing was explicit. As the Prime Minister (then Minister for Social Services) stated in May 2015: “We’re going to track down people who are not doing the right thing, who are owing money to the government and therefore the taxpayers, and we are going to finally get the money back off them”.

The program was expected to save the government $2 billion in the federal budget. That figure — the projected saving — was the measure by which the scheme’s success would be judged. Not the accuracy of the debts. Not the impact on recipients. The saving.

3.2.2 The Methodology

The scheme relied on a process called income averaging. The mechanism was described by the Royal Commission as follows:

· DHS obtained annual income data from the Australian Taxation Office (ATO).

· This annual figure was divided by 26 to produce an averaged fortnightly income.

· If the averaged fortnightly income exceeded what the recipient had declared, DHS inferred that they had been overpaid.

· A debt was raised. The recipient was required to disprove it.

The premise was unsound. As the Royal Commission stated: “The premise for the scheme was unsound because it treated average earnings as though they were actual earnings. But that was most unlikely to be true unless the recipient earned the same amount every fortnight. The very circumstances which caused people to need benefits meant that it was very improbable that many did”.

The scheme applied to debts going back to the financial year 2010–11. Recipients were required to produce evidence of their actual fortnightly income for periods up to six years prior. Many no longer had access to payslips from that period.

3.2.3 The Reverse Onus

The structural core of the scheme was the reverse onus. As Victoria Legal Aid describes it: debts were “imposed on people which they then had to prove they didn’t owe”. The Royal Commission described the process: “The benefit recipient was required to displace what was effectively a presumption of overpayment based on the ATO’s financial year information. In some instances, recipients were required to demonstrate that the averaged ATO figures did not reflect their actual fortnightly income for periods going back years. If the recipient could not or did not do so, the Department proceeded to use its debt recovery powers without any further attempt to substantiate the alleged debt”.

This is not a legal process. It is a power asymmetry embedded in code. The citizen must prove their innocence. The state does not have to prove its case.

3.3 The Warnings That Were Ignored

3.3.1 The Administrative Appeals Tribunal

The AAT began setting aside Robodebts in 2017. As the Royal Commission documents: “The AAT decisions setting aside Robodebts on the ground that evidence of averaging was not capable, without more, of proving a debt”.

DHS disagreed with these decisions. It “ought to have referred them to DSS for consideration of appeals and it certainly should not have continued to use averaging in disregard of what was said in them. However, DHS took the course instead of taking whatever steps were directed by the AAT to rectify the individual cases by obtaining other evidence, but otherwise ignoring the decisions. The fact that Tier 1 AAT decisions were not published made it easier for it to do so”.

The opacity of the tribunal process — the fact that decisions were not published — became a mechanism for ignoring them.

3.3.2 The Clayton Utz Advice

In August 2018, DHS obtained legal advice from Clayton Utz. The Royal Commission states: “The draft advice which the solicitors provided said that it was not permissible to determine a recipient’s entitlement to benefit by averaging employer-reported income from the ATO.” It “should have prompted, if not the immediate suspension of Robodebt averaging, at the least the immediate obtaining of further advice from the Solicitor-General. However, it was never put into final form or acted on”.

3.3.3 The Solicitor-General’s Advice

In September 2019, the Solicitor-General advised that “averaged PAYG data could not support a conclusion as to the amount or existence of a debt in the absence of further evidence that the recipient earned a consistent fortnightly income over periods precisely aligning with the periods declared by the employer to the ATO”.

That advice led to the settlement of the Amato case, with the Federal Court declaring that “averaged PAYG income information was not capable of satisfying a decision-maker of the existence of a debt”.

The state’s own legal advisers had told it the scheme was unlawful. The scheme continued anyway.

3.3.4 The Institutional Checks That Failed

The Royal Commission was explicit about the failure of oversight mechanisms: “Equally disheartening was the ineffectiveness of what one might consider institutional checks and balances – the Commonwealth Ombudsman’s Office, the Office of Legal Services Coordination, the Office of the Australian Information Commissioner and the Administrative Appeals Tribunal – in presenting any hindrance to the Scheme’s continuance”.

Every institution designed to catch the error failed to catch it. Not because they were unaware, but because the architecture of oversight was not designed to interrupt a scheme with political backing.

3.4 The Human Cost

3.4.1 The Numbers

By the time the scheme was abandoned in November 2019:

· Approximately 500,000 debts had been raised.

· $1.73 billion in debts were raised against approximately 433,000 Australians.

· 381,000 individuals were pursued to repay almost $752 million.

· 470,000 debts were declared “insufficient under law” and reduced to zero.

· $1.751 billion in debts were written off.

· $746 million was refunded to approximately 381,000 affected individuals.

3.4.2 The Psychological Harm

The Royal Commission heard evidence of the profound psychological toll. Witnesses described “distress, trauma and thoughts of suicide triggered by false debt notices that were often unexpected and difficult to challenge”.

The Commission’s report includes direct testimony. One witness stated: “I do remember driving home at night just beside myself with worry about this money and thinking I could just drive my car into a tree and make it stop”.

Felicity Button told the Commission: “I felt suicidal for a period of months in 2017 with the ‘lowest point’ being the occasion when ARL (debt collector) debited money from my account. I felt desperate on that day; it was so upsetting that I could not afford to pay for my daughter’s medical expenses and I felt powerless to improve my situation”.

3.4.3 The Deaths

The Royal Commission heard evidence from the mothers of two young men who died by suicide after being pursued for Robodebts. Their stories are told in the report.

Rhys Cauzzo died by suicide on 26 January 2017, aged 28. He had been sent a letter in May 2016 advising that DHS was reviewing his employment income. He told his mother that Centrelink had not explained how the debt had been raised, and that he was not able to provide historical payslips because he could no longer access them.

The Commission concluded: “The Commission is confident that these were not the only tragedies of the kind”.

The Commission also noted that DHS was aware of the risk. Kathryn Campbell, former secretary of DHS, observed in her evidence that “suicide was something that we [at DHS] dealt with frequently”. The Commission stated: “That DHS was aware of this likelihood – that it dealt with suicides frequently – makes the implementation of the Scheme all the more egregious, particularly when there was evidence that they were raising inaccurate debts”.

3.4.4 The Compounding Harm

The scheme inflicted its heaviest toll on those least able to bear it. As Victoria Legal Aid notes: “Robodebt showed us how much damage inadequate government policy design and implementation can cause, how far it can reach, and the compounding harm created for people already under stress”.

One witness stated: “My hope for the future is that Centrelink listens to people when they say something is incorrect and do not abuse their power. I was in a privileged position when my debt was raised. I was employed. There were lots of people still receiving Centrelink who were powerless to do anything about it”.

3.5 The Resolution

3.5.1 The First Class Action

In November 2019, a class action was commenced against the Commonwealth. On 11 June 2021, the Federal Court approved a settlement requiring the Commonwealth to:

· Consent to court declarations that some of its administrative decisions were not validly made

· Not raise, demand or recover from certain group members any invalid debts

· Pay $112 million (inclusive of legal costs) in interest to certain group members

The settlement was in addition to the Government’s promise to refund approximately $751 million to affected individuals and withdraw approximately $1.763 billion in debts based on income averaging.

3.5.2 The Second Class Action (Knox)

In September 2024, Gordon Legal appealed the original settlement. In September 2025, the parties agreed to a new settlement — the Knox settlement — which would see the Commonwealth pay an additional $475 million as compensation to eligible group members.

The Federal Court approved the new settlement on 23 June 2026. The total settlement sum is $548.5 million, comprising:

· $475 million as compensation to eligible group members (minus a 15% funding commission)

· Up to $13.5 million for legal costs

· Up to $60 million for administration costs

The combined total value of the settlement, the amounts repaid, paid as restitution, or abandoned by the Commonwealth is estimated at over $2 billion.

3.5.3 The Apology

On 11 June 2020, the Prime Minister issued an apology for the conduct of the program and stated that he “deeply regretted any hardship that had been caused to people from income compliance activities”.

The apology was made in question time. It was, as one parliamentary speech noted, “grudging” and “completely lacked grace and contrition”.

3.6 The Accountability Outcome

3.6.1 The Royal Commission Referrals

The Royal Commission’s report included a sealed chapter containing referrals of information concerning some persons for further investigation by other bodies. Commissioner Holmes stated: “That in part is intended as a means of holding individuals to account, in order to reinforce the importance of public service officers’ acting with integrity”.

3.6.2 The National Anti-Corruption Commission

The NACC’s investigation — Operation Myrtleford — led by Deputy Commissioner Kylie Kilgour made findings of serious corrupt conduct against two of the six referred persons .

The six were named in March 2026: former prime minister Scott Morrison, and public servants Kathryn Campbell, Renee Leon, Mark Withnell, Serena Wilson, and Catherine Halbert.

The NACC found insufficient evidence to refer any individual to criminal prosecution. The two found to have engaged in serious corrupt conduct were Serena Wilson and Mark Withnell.

3.6.2 The Families’ Response

Jenny Miller, whose son Rhys Cauzzo died by suicide in January 2017, responded to the NACC findings: “It’s hard to describe because I feel like I’ve put aside nine years of my life and gotten nothing out of it at all. Like, no accountability, no justice, no nothing. It really needs to go in front of a judge and jury and for them to be found guilty or not guilty”.

3.7 What the Algorithmic Lens Saw

The Robodebt scheme is the paradigm case because it demonstrates, with precision, what the algorithmic lens sees and what it cannot see.

What the lens saw:

· A discrepancy between ATO annual income and Centrelink declared income

· A potential overpayment

· A debt to be raised

· A savings target to be met

What the lens did not see:

· The casual worker whose income fluctuated week to week

· The recipient who no longer had payslips from six years ago

· The person who could not navigate the online system

· The person who paid the debt because they could not prove their innocence

· The mother who could not afford her daughter’s medical expenses after the debt collector debited her account

· The young man who drove home at night thinking he could drive his car into a tree

The lens showed the discrepancy. The human being was in the gap.

The Royal Commission’s central observation is the structural diagnosis: “It is remarkable how little interest there seems to have been in ensuring the Scheme’s legality, how rushed its implementation was, how little thought was given to how it would affect welfare recipients and the lengths to which public servants were prepared to go to oblige ministers on a quest for savings”.

The quest for savings was the objective. The algorithmic lens was the instrument. The human cost was the externality.

3.8 The Structural Lesson

Robodebt did not fail because the algorithm was faulty. It succeeded — for four years — precisely because the algorithm did what it was designed to do: process discrepancies at scale, raise debts, and recover money.

The failure was external to the system. It came from the Federal Court, which declared the scheme unlawful. It came from the class action, which forced compensation. It came from the Royal Commission, which documented the harm.

The system itself did not self-correct. It could not. The algorithmic lens does not see the human being, and the decision-makers who relied on the lens did not see them either.

That is the paradigm case. That is the lesson. The lens shows what it was built to show. The rest is in the gap.

Claim- Status -Summary

# -Claim -Status

1 -Robodebt used income averaging from ATO data to raise debts -Established

2- The scheme applied a reverse onus, requiring recipients to disprove debts -Established

3- The premise of averaging was unsound because it treated average as actual earnings -Established

4 -The AAT set aside Robodebts from 2017; DHS ignored the decisions -Established

5- Clayton Utz advised averaging was impermissible; advice was not acted on -Established

6- Solicitor-General advised averaged data could not support a debt -Established

7- Institutional checks (Ombudsman, OAIC, AAT) failed to stop the scheme -Established

8 -Approximately 500,000 debts were raised; $1.751 billion written off -Established

9- The scheme caused documented psychological harm including suicide -Established

10 -Two deaths were documented in detail; the Commission was confident there were more -Established

11- The total settlement and refunds exceeded $2 billion- Established

12 -NACC found serious corrupt conduct against two of six referred persons -Established

13- No individual was criminally prosecuted- Established

14- The algorithm saw discrepancies; it could not see human context -Inference

15- The system did not self-correct; external intervention was required- Inference

Chapter 4: The NDIS — Epistemic Injustice and the Datafication of Disability

4.1 The Scheme That Was Promised

The National Disability Insurance Scheme was designed as a radical departure from the institutional model of disability support. The Productivity Commission’s original design, released in 2011, proposed a scheme based on individualised funding, user choice, and control. The person with disability would be the expert in their own life. They would choose their supports. The state would provide the resources, not dictate the terms.

The NDIS was legislated in 2013 and rolled out progressively from that year. By 2026, it supported approximately 761,000 Australians. It is the largest social policy reform in Australia since Medicare.

The scheme was not designed with an algorithmic lens. It was designed with a human lens — planners, assessors, support coordinators who would work with participants to understand their needs and build a plan that reflected them.

That design has been progressively dismantled.

4.2 The Independent Assessment Proposal (2020–2021)

4.2.1 The Rationale

In 2020, the NDIA proposed the introduction of independent assessments — standardised functional capacity assessments conducted by government-contracted healthcare professionals, rather than assessments sourced by the participant from their own treating professionals.

The NDIA’s stated rationale was equity. The current system, it argued, relied on individuals “seeking their own assessments at their own expense, resulting in those who have the capacity to pay for and gather assessments receiving more funds in their plans, on average”. The NDIA estimated that participants were spending approximately $140–$170 million per year on private assessments, “making them out of reach for many”.

The NDIA also argued that the current approach resulted in “inconsistencies and variability in the scope, type and quality of information provided”.

4.2.2 The Resistance

The proposal met immediate and sustained resistance from the disability community.

Disability Representative Organisations (DROs) warned that the assessments would “datafy disability in particular ways” — reducing the person to a numerical score that could not capture the complexity of lived experience.

The academic analysis of the campaign identified three distinct harms:

1. Datafication of disability. The transformation of human experience into “the computer legible ‘language of numbers’“. Algorithmic systems “sort human bodies into categories, often based on the diagnostic categorization of impairments”.

2. Epistemic injustice. The individual’s own knowledge of their body is “undermined by interpretations of reality based on data”. The person with disability is positioned as an unreliable narrator of their own experience.

3. Distributive injustice. Resources are allocated through “algorithmic systems of profiling, classification and risk prediction” that “privilege stable, hegemonic ways of knowing, measuring and living with disability”.

The campaign was successful. The government backed down, citing the “concern and fear” that arose from the expectation of increased automation.

But the victory was temporary.

4.3 The New Framework Planning (2025–2026)

4.3.1 The Return of the Algorithm

In December 2025, Guardian Australia revealed that under a major overhaul of the NDIS to be rolled out from mid-2026, funding and support plans for participants would be generated by a computer program, and staff would have no discretion to amend them.

The details were outlined to NDIA staff in an internal briefing. They had not been made public.

Under the New Framework Planning model:

· An assessment process would feed data into a computer system.

· The system would generate a budget.

· The NDIA delegate’s role would be limited to accepting or rejecting the budget.

· There would be no scope for either the NDIA delegate or the Administrative Review Tribunal (ART) to amend the budget.

Dr Georgia van Toorn, writing for the UNSW Disability Innovation Institute, put it plainly: “In other words, the computer will have the final say”.

4.3.2 The Appeal Rights Problem

Under the new model, the ART would no longer have the authority to alter a person’s plan or reinstate funding. Instead, it would only be able to send the plan back to the NDIA for the agency to conduct another assessment.

Dr George Taleporos, independent chair of Every Australian Counts, stated: “If the NDIA is the only body allowed to reassess a plan, even when a tribunal finds something wrong, then the system stops being fair and accountability is lost”.

The DROs’ joint statement described the changes as “narrow[ing] the grounds on which decisions can be challenged and limit[ing] the scope of the Administrative Review Tribunal”.

4.3.3 The “Robo-Planning” Label

Advocates did not mince words. Senator Jordon Steele-John called the changes “robo-planning” that “risks the lives of NDIS participants”.

Dr Stevie Lang Howson, from Disabled People Against Cuts, described the details as “a nightmare scenario for disabled people”:

“They can’t be serious about allowing assessors who don’t have to have allied health qualifications and a computer system to decide our quality of life. This is a terrifying plan that will hurt disabled people.” 

The NDIA denied that the new process was “automated,” stating that “the assessment will be conducted through a semi-structured interview process”. But the structure — data in, budget out, no discretion, limited appeal — is the structure of automated decision-making, regardless of the label.

4.4 The Evidence of Harm

4.4.1 The Reassessment Data

The harm is not hypothetical. It is already being documented.

A survey by Children and Young People with Disability Australia (CYDA) in May 2025 found that of respondents who had received an eligibility reassessment letter:

· 22% were removed from the NDIS 

· 26% had their funding reduced or services cut 

Among parents of children aged nine or under:

· 36% were removed from the NDIS 

· 14% had their funding reduced 

A separate CYDA webinar reported a 90 per cent removal rate for children and 10 per cent for adults.

4.4.2 The I-CAN Tool

The NDIA has developed the I-CAN tool — a functional assessment instrument designed to categorise participants and allocate budgets.

A submission to the Joint Standing Committee on the NDIS described the tool’s logic:

“The use of I-CAN as a tool to categorise people to allocate them a budget reduces participants to actuarial risks. This fiscal reductionism exemplifies an ableist worldview where disabled lives are framed primarily as costs to be controlled.” 

The submission warned that the tool “could easily be used to rationalise placements into” group homes and congregate care — “creating support packages that make institutional care appear administratively efficient”.

The critique of the implementation strategy was equally sharp:

“The NDIA is centralising discretion while claiming to decentralise it. The messaging claims that I-CAN reduces subjectivity by removing ‘variability between professionals.’ However, what is actually occurring is the centralisation of interpretive power within the NDIA’s pricing and planning algorithms.” 

The submission noted that the NDIA had not provided “a published clinical governance framework for I-CAN use,” “practice guidance for how assessors should interpret ambiguous or fluctuating needs,” or “training in trauma-informed interviewing”.

4.5 The Failure of Recognition

4.5.1 Three Forms of Misrecognition

A 2024 study by UNSW researchers identified three distinct forms of failed recognition that people with disability experience when interacting with automated systems:

1. Failure to recognise the diversity of disability experience. Algorithmic systems “sort human bodies into categories, often based on the diagnostic categorization of impairments.” But “disability is a contested and nuanced concept,” and “the experience of disability is not easily represented by algorithmic quantification”.

One participant described the experience:

“The complete questionnaire didn’t fit at all … none of these questions fitted. So, I would just get very cranky … I felt disregarded.” 

2. Epistemic failure. The system does not acknowledge “the particular kinds of knowledge held by people with disability.” The person’s own knowledge of their body is subordinated to the system’s interpretation.

3. Ontological failure. The encounter between the disabled person and the automated system is a failure of being seen as a person. The system does not recognise the person as a subject with their own experience. It recognises them as a data point.

4.5.2 The Data That Cannot Be Captured

The DROs’ statement is precise about the risk:

“If historical data has under-represented people with a psychosocial disability, intellectual disability, neurodevelopmental disability, complex communication needs, people with fluctuating, multiple or complex disabilities, First Nations people, and other intersectional and marginalised communities, ADM will reproduce and amplify those gaps.” 

“Algorithms built from ‘average’ cases consistently fail those whose experiences sit at the margins.” 

The people whose disabilities are “complex, fluctuating or not easily captured through standardised assessments” are the ones most likely to have their funding cut or supports misjudged.

4.6 The Parallel with Robodebt

The structural parallel with Robodebt is not rhetorical. It is precise.

Robodebt- NDIS New Framework

Algorithm raised debts based on averaged income -Algorithm generates budgets based on standardised assessment

Reverse onus: recipient must disprove- Limited appeal: participant can request reassessment by the same agency

AAT set aside decisions; DHS ignored them ART can only refer back to NDIA; cannot vary the plan

470,000 unlawful debts -240,000 projected to transition off the scheme 

Harm concentrated on vulnerable Harm concentrated on complex, fluctuating, marginalised disabilities

The structural logic is identical. The algorithm processes what it was designed to process. The human being is in the gap. And the accountability architecture has been adjusted to prevent the gap from being closed.

4.7 The Structural Capture

What is happening to the NDIS is not a failure of the scheme. It is the application of the neoliberal market logic documented in Chapter 1.

The scheme was designed as a market: participants choose providers, funding follows the person. But markets require measurement. Measurement requires instruments. Instruments produce data. Data feeds the algorithm.

The Productivity Commission’s original design was based on the premise that the person is the expert. The algorithm is based on the premise that the system knows better. The person’s knowledge — their lived experience, their self-assessment, their knowledge of their own body — is discredited in favour of standardised data.

That is epistemic injustice. And it is not an accident. It is the logical consequence of designing a human service as a market and then automating the market’s measurement.

The person with disability is not the customer. The person is the data source. The algorithm processes them. The budget is generated. The NDIA delegate accepts or rejects. The ART refers back. The computer has the final say.

4.8 What the Lens Does Not See

The NDIS algorithm measures functional capacity. It cannot measure:

· The fluctuating nature of complex disability

· The impact of poverty, discrimination, and social context

· The value of a relationship with a trusted support worker

· The trauma of repeated assessment

· The fear of losing support

· The person’s own knowledge of what they need

The algorithm sees what it was designed to see. The person is in the gap.

Dr van Toorn’s conclusion is the structural diagnosis:

“Just like with robodebt, the system will inflict its heaviest toll on the communities least able to bear it.” 

Claim- Status- Summary

#- Claim- Status

1 -NDIS designed as individualised funding with user choice- Established 

2- NDIS supports ~761,000 Australians- Established 

3- Independent assessments proposed in 2020 and withdrawn- Established 

4 -New Framework Planning generates budgets via computer program -Established 

5 -NDIA delegate limited to accept/reject; no discretion to amend -Established 

6- ART cannot vary plan; can only refer back to NDIA- Established 

7 -22% of reassessed removed; 26% had funding cut -Established 

8 -90% removal rate for children in reassessments- Established 

9- I-CAN tool reduces participants to actuarial risks -Established 

10- DROs warn ADM reproduces and amplifies gaps- Established 

11 -Algorithms built from “average” cases fail marginalised -Established 

12- 240,000 projected to transition off scheme -Established 

13- Structural parallel with Robodebt- Inference

14 -Algorithm sees functional capacity; cannot see human context -Inference

Chapter 5: Aged Care — The Integrated Assessment Tool and the Outsourcing of Care

5.1 The Royal Commission and Its Aftermath

The Royal Commission into Aged Care Quality and Safety delivered its final report in March 2021. It was the culmination of two years of hearings, thousands of submissions, and documented evidence of systemic neglect. The Commissioners found that the system was characterised by “inadequate funding, variable provider governance and behaviour, absence of system leadership and governance, and poor access to health care”.

The Commission’s central finding was damning: “The current state of Australia’s aged care system is a predictable outcome of these measures to limit expenditure and ignore the actual cost of delivering aged care”. The Commissioners recommended a fundamental shift — a new rights-based Act, entitlement to care based on need, and a financing model that would provide predictable and adequate funding.

The Commission also recommended a single assessment system to replace the fragmented assessment processes that had characterised the system.

What followed was not the implementation of the Commission’s vision. It was the outsourcing of its implementation to the very interests the Commission had criticised.

5.2 The Single Assessment System and the IAT

5.2.1 The Reform

On 1 July 2024, the Integrated Assessment Tool (IAT) was introduced as the standard tool for determining eligibility for government-subsidised aged care services. The IAT replaced the National Screening and Assessment Form (NSAF) and was designed to streamline assessment across home support, comprehensive in-home care, and residential aged care.

On 9 December 2024, the Single Assessment System workforce officially replaced the previous Regional Assessment Service (RAS), Aged Care Assessment Teams (ACATs), and independent AN-ACC assessors. The stated intention was to ensure that older Australians could remain with the same assessment provider as their needs evolved, improving continuity of care and reducing wait times.

The IAT is a structured digital assessment tool. Assessors enter information about the older person’s needs across a range of domains: carer profile, medical and medications, behaviour, social, cognition, psychological, home and personal safety, financial or legal, and support considerations. The tool then produces a classification that determines the level of funding.

5.2.2 The Algorithm

The critical development came with the Support at Home program. According to departmental guidance, the IAT classification algorithm is now a mandatory step in all aged care assessments. Assessors must complete all mandatory questions “to allow the IAT algorithm to run upon finalisation”.

The guidance states that assessors “can accept or override the recommendation but must provide justification”. This appears to preserve discretion. But the operational reality, as reported by assessors, is different.

An assessor told the ABC: “I am forced to sign off on decisions I don’t agree with”. The tool, the assessor said, “frequently assigns low levels of care to people with complex or high needs, leaving clinicians powerless”.

The assessor’s account describes a structural feature of the algorithm: the clinician conducts the home visit, asks the questions, and enters the data. But the classification is produced by the algorithm. The clinician’s role is reduced to data entry and sign-off.

5.2.3 The Opacity

The classification logic of the IAT has not been released. A 2026 analysis noted that “despite requests for the technical specifications, and for the identity of the team that designed the classification logic, details have not been released”. The health department insists the tool “is not artificial intelligence (AI)” but “a rules-based classification algorithm”. But as the analysis noted: “whether a system involves AI or not is beside the point. Deeper issues sit underneath the technology”.

Those deeper issues are opacity, discretion, and accountability. The people affected cannot see how the system works. The professionals using it cannot exercise judgement. And no one can be held to answer for its decisions.

5.3 The Outsourcing of Assessment

5.3.1 The Privatisation

One of the most consequential changes has been the partial privatisation of aged care assessments. Contracts worth approximately $1.2 billion were awarded to private providers, replacing the state and territory public health clinicians who had conducted assessments for decades.

The Australian Medical Association had warned against this precisely. In its submission on the new in-home aged care program, the AMA stated: “It is disappointing to see that assessment privatisation is back on the agenda, almost three years after the process was abandoned”. The AMA argued that “the aged care assessment service for older people with complex health and aged care needs must remain with State/Territory health services”.

The AMA’s concern was structural: “There is real fear among the medical profession that a market approach will result in exclusion of medical specialists from the assessments, resulting in further fragmentation of care for older people, leading to increased and prolonged hospital stays for many of them”.

The fear was justified. The outsourcing of assessments created a system where the clinical expertise that had been embedded in ACATs — linked to local hospital networks — was replaced by a contracted workforce with less connection to the health system.

5.3.2 The Consultancy Layer

The outsourcing extends beyond assessments. The Aged Care Quality and Safety Commission has “outsourced nearly $600 million in AusTender contracts that have included advice and shadow workforce” since 2021. One submission to the Independent Governance Advisory Committee described the Commission as appearing “more as a brokerage than a regulator”.

The Department of Health and Aged Care is “reliant upon outsourcing for essential services such as auditing, and financial advice modelling”. Its advisors include KPMG, Nous Group, Stewart Brown, PwC, HealthConsult, Accenture, and “a stack of niche advisors servicing aged care”.

The structural consequence is that the information that reaches decision-makers is pre-filtered by the consultancy layer. The consultancies are accountable to the department, not to the older person. They optimise for contract deliverables, risk management, and cost efficiency. Human impact is not their metric.

5.3.3 The Failure of Implementation

The Royal Commission’s recommendations were substantially ignored. As one submission to the Independent Governance Advisory Committee stated: “It totally confounds the hoards of advocates who participated in thousands of hours of consults, webinars, conferences and the writing of submissions, that the work of the Royal Commission into Aged Care Quality and Safety was to the main ignored”.

The submission identifies the structural mechanism: “The Labor government fearful of electoral loss opted to follow outsourced advice rather than consider that aged care should be directed and where possible managed by government”. The Aged Care Task Force — chaired by the Minister and composed largely of members “from the very aged care system that led to the Royal Commission” — took precedence over the recommendations of the independent Commission.

The result is a system that has been restructured according to market logic rather than the rights-based model the Commission recommended.

5.4 The Human Cost

5.4.1 The Waiting Lists

The delays are documented and severe.

As of March 2025, 87,597 people were waiting in the National Priority System for a Home Care Package at their approved level. This represented a 184 per cent increase over two years — from 30,839 people in March 2023.

The estimated wait times ranged from 6 to 15 months depending on the package level, with medium priority applicants waiting 8 to 13 months. At the 90th percentile, older people could wait up to 424 days for a Home Care Package.

By October 2025, the Minister for Aged Care confirmed the waiting list had grown to over 120,000 people waiting an average of ten months, with some waiting 15 to 18 months. A separate analysis estimated that around 200,000 people need home care and are not receiving it.

5.4.2 The Assessment Backlog

The privatisation of assessments created its own backlog. More than 116,000 people are currently waiting for a home care assessment. While the government advises the median wait time is 23 days, families and carers describe a vastly different experience.

The ABC reported being “inundated with emails” following a 7.30 report. One couple in their 70s were told it would take “nine to 10 months” just to be assessed. Another woman waited eight months for her father’s reassessment — only to be contacted two months after he had died. An elderly man said he had simply given up.

5.4.3 The Cost Spiral

The submission to the Independent Governance Advisory Committee documents the cost escalation:

· In 2021, pressure socks cost $50 per pair. By 2025, the same brand and type cost $100.

· The agreed Residential Accommodation Deposit (RAD) was $450,000 in 2020. By 1 January 2025, it was $750,000.

· A home walker cost $350 in 2019. By 2025, it was $650–$750.

· The Aged Care Budget was $17.7 billion in 2021–22. By 2024–25, it was $36.2 billion.

The submission states: “Greater level of privatisation more than doubled the cost in four years without significant increase in participant numbers but with a decrease in number of services available topped up with increased co-contribution”.

5.4.4 The Co-Contributions

Under Support at Home, Non-Clinical Care Contributions (NCCC) were introduced. Many essential in-home services — including showering, dressing, cleaning and shopping — now attract co-payments. These are not discretionary extras. They are fundamental supports that underpin health, dignity and quality of life.

The government has since announced that from 1 October 2026, personal care will be fully funded. But the other co-contributions remain.

The “no worse off” promise — that the 250,000 people already in the system would be grandfathered — has not held in practice. Some providers, capped on management fees, responded by sharply increasing hourly service rates — in some cases by 30 to 40 per cent. The result is fewer hours of care each week, even though funding levels have not increased.

5.4.5 The Delayed Discharge

The consequences extend into the hospital system. Some states estimate that around 15 per cent of their hospital beds are taken up by older people who should be in a residential aged care bed. The Minister acknowledged “delays in getting people, out of hospital and into those aged care beds”.

The structural connection is direct: if the assessment and allocation system is delayed, older people remain in hospitals because there is nowhere else for them to go. The hospital bed becomes the waiting room for the aged care system.

5.5 What the Lens Does Not See

The IAT measures functional capacity. It asks about mobility, cognition, daily living, and social circumstances. It produces a classification that determines funding.

What the IAT cannot measure:

· The fear of an older person who has been told they must wait ten months for care.

· The dignity of a person who must ask their daughter to help them shower because the assessment classification did not provide for a support worker.

· The exhaustion of a carer who resigned from their job because the package never arrived.

· The grief of a family contacted about a reassessment two months after their father died.

· The person behind the data.

The algorithm processes what it was designed to process. The human being is in the gap.

5.6 The Structural Parallel

The pattern is the same as Robodebt and the NDIS.

Feature Robodebt – NDIS Aged Care

Algorithm determines outcome -Income averaging Budget generation IAT classification

Human discretion limited DHS delegate- NDIA delegate Assessor cannot override

Appeal mechanism restricted AAT decisions ignored- ART cannot vary Limited review

Human impact external to system Royal Commission -DRO campaigns Royal Commission

Harm concentrated on vulnerable Welfare recipients -Complex disabilities Frail elderly

Documentation comes from outside Courts, RC Advocates ABC, submissions

The structural logic is identical. The state has outsourced the delivery of care to a market. The market requires measurement. The measurement is performed by an algorithm. The algorithm sees what it was designed to see. The human being is in the gap.

5.7 The Privatising of Morality

The submission to the Independent Governance Advisory Committee uses a phrase that captures the structural transformation precisely: “The privatizing of morality”.

The phrase describes what happens when a state outsources not merely the delivery of services but the judgment about who deserves what. The assessment of need becomes a market transaction. The allocation of care becomes a contract. The older person becomes a data point in a system designed to optimise for cost, risk, and efficiency.

The Royal Commission recommended a rights-based system. The government implemented a market-based system. The difference is not semantic. It is structural. A rights-based system sees the older person as a subject with entitlements. A market-based system sees them as a consumer with purchasing power, or as a cost to be managed.

The algorithm is the instrument of the market. It measures what the market needs to measure. It cannot see what the rights-based system would have required it to see: the person, in their fullness, with their needs, their dignity, and their fear.

Claim -Status- Summary

# -Claim- Status

1 -Royal Commission found system characterised by inadequate funding, poor governance -Established 

2- Commission recommended single assessment system -Established 

3 -IAT introduced 1 July 2024 as standard assessment tool- Established 

4- IAT classification algorithm is mandatory; assessors can accept or override -Established 

5 -Assessors report being forced to sign off on decisions they disagree with -Established 

6- IAT classification logic has not been released- Established 

7- Assessments partially privatised; contracts worth ~$1.2 billion -Established 

8- AMA warned privatisation would fragment care- Established 

9 -Aged Care Quality Commission outsourced ~$600m in contracts -Established 

10- 87,597 people waiting for HCP as of March 2025; 184% increase -Established 

11 -Waiting list grew to 120,000 by October 2025- Established 

12 -~200,000 people need home care and are not receiving it- Established 

13- 116,000 waiting for assessment- Established 

14- Aged care budget doubled from $17.7b to $36.2b without increased participants -Established 

15- 15% of hospital beds occupied by older people awaiting aged care -Established 

16- Algorithm sees functional capacity; cannot see human context- Inference

17 -Structural parallel with Robodebt and NDIS- Inference

18 -Privatising of morality describes the structural transformation -Inference

Chapter 6: Veterans — The Royal Commission and the System “Geared to Frustrate”

6.1 The Commission That Named the System

The Royal Commission into Defence and Veteran Suicide delivered its Final Report in September 2024. It was the culmination of three years of hearings, thousands of submissions, and the testimony of families who had lost sons and daughters to suicide. It produced 122 recommendations across seven volumes.

The Commission’s central finding was structural. As Commissioner Peggy Brown stated in her closing address: “This is not a problem that Defence or DVA can be trusted to solve by themselves. We have seen too often that Defence’s default position is to defend Defence, and too much evidence of ‘busy work’ at both Defence and DVA giving the appearance of implementing change, when in reality they have only served to perpetuate old ways of operating.” 

The Commission’s task was not to find individual villains. It was to identify the systemic failures that contribute to suicide risk. And what it found was a department that had become, in the words of one parliamentarian, “a bureaucratic machine, crushing the people it should be helping”.

6.2 The Claims Process as a Stressor

6.2.1 The Backlog

The Commission identified the claims process as a key stressor. Chapter 25 of Volume 5 recommended “significantly increasing the percentage of claims determined within 90 days”.

The scale of the problem was documented in the Commission’s evidence. As one veteran advocate wrote in a submission: “Average initial-liability wait times remain 303 days as at April 2025”. The advocate’s own claim had been running for 578 days and counting.

The human cost of the delay was described in the Commission’s report. A former Secretary of DVA acknowledged that the department “relies on a reactive model where the veteran has to come forward to access support”. The Commission noted that DVA “does not use specific methods to understand the experiences of veterans who have not initiated engagement”.

This is the algorithmic lens in operational form. The system waits for the veteran to lodge a claim. The claim is processed through a system designed to manage volumes, not to see the person. The veteran experiences silence, uncertainty, and repeated handoffs.

6.2.2 The Evidence Problem

The Commission heard evidence that DVA assessors were applying an internal rule that always required corroborative evidence, even though the legislation itself did not. This policy was identified by the Royal Commission into Institutional Responses to Child Sexual Abuse in 2017, which found that the rule “may have the effect of denying claims that should have been accepted”.

The burden was placed on the veteran. They had to prove their injury was service-related. They had to produce evidence that the department could have obtained itself. And while they waited, their health deteriorated.

6.2.3 The Fragmentation

The Commission repeatedly identified fragmentation across transition, claims, health care and veteran support. Veterans were “transferred between areas, repeatedly explained their circumstances and struggled to identify who was responsible”.

The Commission recommended that every veteran with an active claim or complex support need should have “one named DVA contact from commencement to resolution”. The rationale was explicit: “DVA’s organisational complexity should be managed inside DVA rather than imposed on an injured or vulnerable person as another problem they must solve”.

The Commission also noted that DVA had established a two-day callback standard, yet senior officials responsible for the area could not explain how compliance was measured. As the Commission stated: “For a vulnerable veteran an unanswered call can become a point of crisis, so a safeguard that is neither monitored nor enforced is not a meaningful safeguard”.

6.3 The Human Cost

6.3.1 The Numbers

Over the past 10 years, an average of 78 serving or ex-serving ADF members have died by suicide each year. That is three lives lost every fortnight.

The AIHW’s 2025 report found that suicide remains the leading cause of death for younger and ex-serving members who separate from the ADF involuntarily, with higher rates than those who separate voluntarily.

The Commission found that only 29 per cent of ex-serving members who died by suicide between 2002 and 2021 were DVA clients, meaning 71 per cent were not. The majority of those who died had not accessed the support system at all.

6.3.2 The Witnesses

The Commission heard from families whose sons had died. Their testimony is in the record.

Ryan Goodwin died by suicide after transitioning from the Army. His father told the Commission: “Ryan experienced occupational and social impairment and displayed symptoms such as the following. He lost his passion for reading books, became very introverted in conversations and lost contact with his comrades. He had a disrupted sleep pattern and the memories of his service prevented him from sleeping. He displayed inappropriate behaviour, and became angry, frustrated and verbally aggressive.” 

The family had tried to get Ryan help. They got him to an RSL advocate. But as his father wrote: “I believe dealing with DVA causes more mental health problems than the original incidents sometimes”.

Michael died by suicide after his psychiatrist noted he was talking about suicide but “at the end we were laughing about that.” Michael’s mother told the Commission: “That day he had done one of two things: told his psychiatrist that he was suicidal, and then went straight to the lawyer’s office and said ‘I want to withdraw my claim.’ So the biggest thing was he didn’t want me to have to pay. It was quite a large sum of money, it was $10,000.” 

The system had become so difficult that he withdrew his claim to spare his mother the cost. And then he died.

6.3.3 The Advocates’ Accounts

The Commission heard from Lance Johnson OAM, a former acting Deputy Commissioner for DVA who became an advocate. His testimony is precise:

“Even today, I am still coming across situations where DVA does not give the necessary priority to an obvious situation of risk.” 

He described a case where a veteran with an accepted knee injury applied for reassessment. The claim was rejected because the doctor had used “slightly different wording to state the condition.” The veteran was unemployed. The advocate wrote: “There was no sense testing. No element of reasonableness. This has placed my veteran in a situation where he is at risk of suicide – he could commit suicide any day given the pressure this puts him under.” 

The advocate’s conclusion was structural: “DVA will not apply judgment or actively consider things and advance things. Many staff members within DVA don’t think and are policy driven rather than veteran-centric.” 

6.4 The Algorithmic Lens in Operation

6.4.1 The AI Transparency Statement

DVA published an AI Transparency Statement in April 2026. It states that the department uses AI for “text analysis, including keyword matching” in the MyService portal, and that this technology “does not make decisions about individuals”.

The statement also notes that DVA intends to commence a voluntary, consent-based trial of a Prototype AI Claims Document Reader in partnership with RSL Queensland. The prototype is designed to “assist staff by quickly identifying relevant information in lengthy claims documents.” All claims’ decisions remain with human delegates.

The department’s framing is careful: “DVA does not use AI to make decisions without human involvement.” 

But the structural question is different. The question is not whether the algorithm makes the decision. The question is what the algorithm sees and what the human delegate sees.

6.4.2 The Business Rules Engine

A DVA submission to the Joint Committee of Public Accounts and Audit reveals a more significant development. DVA’s Request for Information process included a use case for “a scalable business rules engine, that has machine learning (ML) capability. The solution must be able to ingest legislation, policy documentation, or website content and convert the data into machine-readable code for assessing claims and eligibility.” 

The department acknowledges that its legislative framework “enables the use of computer-based decision-making on certain claims, but only to accept liability for those conditions.” The VETS Bill does not change this.

This is the architecture of automation. The legislation is converted into code. The code assesses claims. The human delegate is retained for adverse decisions. The system is designed to streamline the process.

But what does the system see? It sees the legislation, the policy, the documentation. It applies the rules. What it cannot see is the veteran who has been waiting 578 days, who has been asked twice for the same evidence, who has been told his claim is “not within the norm” and rejected without explanation.

6.5 The Structural Parallel

The pattern is the same as Robodebt, the NDIS, and Aged Care.

Feature Robodebt NDIS Aged Care Veterans

Algorithm determines outcome Income averaging Budget generation IAT classification Business rules engine

Human discretion limited DHS delegate NDIA delegate Assessor cannot override Delegate sees pre-processed data

Appeal mechanism restricted AAT decisions ignored ART cannot vary Limited review 303-day average wait

Human impact external to system Royal Commission DRO campaigns Royal Commission Royal Commission

Harm concentrated on vulnerable Welfare recipients Complex disabilities Frail elderly Veterans with PTSD

Documentation comes from outside Courts, RC Advocates ABC, submissions Families, advocates

The structural logic is identical. The state has designed a system to process claims. The system requires standardised inputs. The inputs are generated by the veteran. The veteran is the data source. The system processes the data. The human being is in the gap.

6.6 The Closing Address

Commissioner Peggy Brown’s closing address is the structural diagnosis, stated in her own words:

“Real time monitoring and honest, open reporting of progress and true accountability will also be critical, aided by external oversight. No matter how sincere their efforts, this is not a problem that Defence or DVA can be trusted to solve by themselves.” 

“There must be a continued drive for reform, to analyse results and to innovate, and a true commitment to tackling the many issues contributing to Defence and veteran suicide.” 

“History will not only judge this current government but all future governments, as well as state and territory governments, and everyone else with the ability to change the future in this space.” 

The Commission drew a line in the sand. The question is whether the system can see across it.

6.7 What the Lens Does Not See

The claims processing system measures liability. It measures entitlement. It measures processing time and backlog.

What it cannot measure:

· The veteran who has stopped answering calls because the system has defeated him.

· The family who has given up because the paperwork is too much.

· The 71 per cent who never became DVA clients and died anyway.

· The veteran who withdrew his claim to spare his mother the cost, and then took his own life.

· The person behind the claim.

The system processes claims. The veteran is in the gap.

Claim -Status- Summary

# Claim Status

1- Royal Commission delivered 122 recommendations -Established 

2 -Commission found DVA cannot be trusted to solve the problem alone- Established 

3- Claims process identified as key stressor; 303-day average wait- Established 

4- DVA relies on reactive model; does not identify non-clients -Established 

5 -Veterans experience fragmentation and repeated handoffs -Established 

6- DVA callback standard not monitored- Established 

7 -78 ADF members died by suicide annually over past decade- Established 

8- 71% of ex-serving members who died by suicide were not DVA clients -Established 

9- DVA uses AI for text analysis in MyService -Established 

10- DVA exploring business rules engine for claim assessment -Established 

11 -Advocates report DVA rejects claims on technicalities- Established 

12- System sees liability; cannot see human context- Inference

13- Structural parallel with Robodebt, NDIS, Aged Care -Inference

Chapter 7: The NSW Mental Health System — Financial Management Orders and the Pipeline to the State Trustee

7.1 The Order That Outlives the Admission

A Financial Management Order (FMO) is a legal instrument that transfers control of a person’s financial affairs to the NSW Trustee and Guardian. It can be made by the Mental Health Review Tribunal (MHRT), the Supreme Court, or the Guardianship Division of NCAT.

The MHRT’s power is specific: it can only make an FMO for a person who is presently an inpatient at a mental health facility. The person may be a voluntary patient, an involuntary patient, or a detained assessable person.

The hearing is an informal meeting. The assessment is made by a Tribunal panel comprising a lawyer, a psychiatrist, and a member with qualifications such as social work or psychology.

The critical structural feature is this: an FMO continues even after the person ceases to be a patient in a mental health facility. If it is not an interim order, it will only come to an end if it is revoked or successfully appealed against.

The admission is temporary. The order is not.

7.2 The Fee Structure

The NSW Trustee charges fees for managing a person’s estate. The current fee structure under the 2026 Regulation is as follows:

Fee Amount

Establishment (one-off) 1% of estate value, capped at 30 fee units

Annual management 1.4% of estate value, capped at 200 fee units

Account keeping 0.1 fee units per account per month

Investment (common fund) 0.35% per annum of investment value

The management fee is calculated on the chargeable assets — which excludes the principal place of residence and assets for personal use such as a motor vehicle and furniture.

The NSW Audit Office found that the average estate value for direct financial management clients over the past three financial years was approximately $192,500, and the average management fee charged per client per annum was around $1,624 . On average, direct financial management clients paid approximately $1,900 per annum in total fees.

The audit also found that NSW Trustee is the only Australian public trustee to not charge the management fee on a sliding scale of different asset values. Most other jurisdictions do not charge a one-off establishment fee, nor an account keeping fee.

7.3 The Audit Findings: What the System Does Not Do

The NSW Audit Office’s 2023 report, Managing the affairs of people under financial management and/or guardianship orders, found that NSW Trustee and Guardian is delivering services in line with its broad legal authority, but does not have sufficient oversight to ensure that its services are consistent with legislative principles.

The key findings:

· It does not track the actual costs of service delivery, the quality of services, or client experiences. Key findings from previous reviews remain unresolved.

· Government funding has not kept pace with the growth in clients.

· There is a risk that some fee-paying clients are unknowingly subsidising others.

· Gaps in monitoring and IT system constraints create a risk that service quality issues will not be addressed.

The audit recommended that NSW Trustee and Guardian:

· Broaden governance to include input from people with lived experience.

· Implement mechanisms to seek feedback from clients.

· Assess staff competency and implement regular training in effectively serving clients with disability, dementia, mental illness, cognitive impairments and other factors relevant to decision-making incapacity.

· Improve data collection and monitoring to track performance, costs, and client outcomes.

7.4 The Self-Reliance Deficit

The audit found limited evidence that NSW Trustee and Guardian staff encourage clients to be self-reliant in matters relating to their personal, domestic and financial affairs, as outlined in the legislative principles.

Under section 71 of the NSW Trustee and Guardian Act 2009, financial managers have discretion to authorise a client to manage part of their own estate as a means of developing their decision-making capacity. This is known as a “section 71 authority.”

The data is stark: only 610 clients out of 12,556 direct financial management clients (five per cent) had section 71 authorities as at 30 June 2022.

The audit noted that NSW Trustee and Guardian established a dedicated team to focus on growing the number of people given section 71 authorities. But it is a small team of four people for a consumer base of at least 12,400 direct financial management clients per annum.

The audit also found that NSW Trustee and Guardian does not provide direct financial management clients with the ability to independently access real-time information about their available funds, expenditure, and budget. This makes it difficult for clients to be self-reliant and independently track their spending.

The structural consequence: the system is designed to manage the person’s affairs for them, not to support them in managing their own affairs. The legislative principle of self-reliance exists on paper. In practice, it is rarely applied.

7.5 The Frontline Staffing Problem

The audit documented the resource constraints that shape the system’s operation.

In 2021–22, NSW Trustee and Guardian modelled the average time spent by frontline staff with clients. The modelling estimated that frontline staff were spending less time each week on average in 2020 compared to 2011. Without additional investment, this would drop to an average of 14 minutes per week per client for guardianship and 11 minutes per week per client for direct financial management in 2026.

The audit also found that inadequate IT systems have made it difficult for staff to record and readily obtain information about individual clients. The systems do not automatically flag or prompt staff and supervisors to check that client records are complete and accurate, and that decisions are in line with relevant client information and the legislative principles.

The consequence is structural: the human judgment that the system requires cannot be exercised because the time and information are not available. The system operates on a default mode of processing, not deliberation.

7.6 The Communication Bottleneck

The hospital environment imposes its own restrictions on the person subject to an FMO.

NSW Health policies on personal electronic devices in mental health units vary by Local Health District. The South Eastern Sydney Local Health District’s Business Rule states that consumers “should have access to their personal mobile phone or device in accordance with least restrictive care,” and that “limited access and restriction should only be considered where a risk has been identified”.

However, where a risk has been identified, the device must be “entered in the property register and stored securely by nursing staff according to local protocols”. The restriction is to be “reviewed on a shift-by-shift basis”.

The note at the end of the Business Rule states: “All inpatient units should have access to a designated public telephone. Wherever possible this should be in an area that promotes and protects privacy”.

The structural effect is that all outgoing communication is mediated through a system controlled by the hospital. If the person’s mobile phone is restricted, they must use a public telephone — typically located in a shared area, with limited privacy, and subject to availability.

The person who is subject to an FMO, and who may wish to contest it or seek legal advice, is dependent on the hospital’s communication infrastructure to do so. The hospital is the institution that triggered the FMO process in the first place.

7.7 The Appeal Pathway

A patient can appeal against an FMO made by the MHRT. The pathways are:

· Appeal to the Supreme Court against a financial management order made by the Tribunal.

· Appeal to NCAT against an order made by the Tribunal or a Magistrate.

An appeal to NCAT may be made by the person to whom the order relates or any other person who was a party to the proceedings.

For a review of the order, the following persons are entitled to apply for revocation or variation:

· The protected person concerned.

· The NSW Trustee.

· The manager of the estate.

· Any other person who, in the opinion of the Tribunal, has a genuine concern for the welfare of the protected person.

The Tribunal must vary, revoke, or confirm the order on review. It can revoke only if :

· It is satisfied the person is capable of managing their affairs, or

· It considers revocation is in the person’s best interests (even if they remain incapable).

The legal architecture exists. The structural barriers to using it — communication restrictions, lack of independent advice, the person’s own cognitive state at the time of the hearing — are the practical obstacles.

7.8 What the Lens Does Not See

The FMO system measures incapacity. It measures the person’s ability to manage their financial affairs. It produces an order that transfers control to the NSW Trustee.

What it cannot measure:

· The person’s fear of losing control over their own money.

· The shame of having to ask for money from a system that holds their assets.

· The frustration of being unable to access real-time information about their own funds .

· The isolation of having their phone restricted in hospital, with only a public telephone available.

· The person behind the assessment.

The system processes the order. The person is in the gap.

7.9 The Structural Parallel

The FMO pipeline shares the structural features of the other systems documented in this work.

Feature -Robodebt -NDIS -Aged Care- Veterans -NSW FMO

Trigger Income discrepancy- Plan reassessment -Assessment Claim -Inpatient admission

Assessment Algorithm- Computer program- IAT Business rules- Tribunal panel

Human discretion limited DHS delegate- NDIA delegate Assessor- Delegate Panel assessment

Appeal mechanism restricted -AAT ignored -ART cannot vary -Limited review- Long waits- Court/NCAT

Human impact external -Royal Commission- DRO campaigns -Royal Commission -Royal Commission -Audit Office

Harm concentrated on vulnerable Welfare recipients -Complex disabilities -Frail elderly Veterans with PTSD -Mental health inpatients

Self-determination deficit- Reverse onus- Limited appeal- Limited override- Reactive model 5% s71 authorities

The structural logic is identical. The state has a process for managing people it deems incapable. The process requires assessment. The assessment is performed by a panel or tool. The outcome is control. The person’s ability to contest the outcome is structurally constrained.

7.10 Conclusion: The Pipeline

The Financial Management Order is not an isolated legal instrument. It is the entry point to a pipeline.

The person is admitted to a mental health facility. The Tribunal makes an FMO. The NSW Trustee takes control of the estate. The order continues after discharge. The person’s ability to have the order revoked depends on their capacity, their access to legal advice, and their ability to navigate the appeal process.

The audit found that the system does not encourage self-reliance. It found that staff have limited time. It found that the IT systems do not prompt them to apply the legislative principles. It found that only 5% of clients have section 71 authorities that would allow them to manage part of their own affairs.

The system is not designed to restore autonomy. It is designed to manage assets. The person is the object of that management.

That is the pipeline. That is the lens. The person is in the gap.

Claim Status Summary

# -Claim- Status

1 -MHRT can only make FMO for current inpatient- Established

2- FMO continues after discharge- Established

3- MHRT cannot appoint private manager; only NSW Trustee- Established

4- Fee structure: 1.4% management, capped -Established

5 -Average annual fee ~$1,900- Established

6 -NSW only public trustee without sliding scale- Established

7- Audit: insufficient oversight for legislative principles- Established

8 -Audit: does not track costs, quality, client experiences -Established

9- Audit: fee-paying clients may subsidise others -Established

10 -Only 5% of clients have section 71 authorities -Established

11- NSW Trustee does not provide real-time fund access- Established

12 -Frontline time projected to drop to 11 min/week/client -Established

13- IT systems do not flag incomplete records -Established

14 -Phone restrictions documented in NSW Health policy -Established

15 -Appeal pathways exist via Supreme Court and NCAT -Established

16 -System manages assets, not autonomy -Inference

17- Person is object of management, not subject- Inference

18- Structural parallel with Robodebt, NDIS, Aged Care, Veterans- Inference

Chapter 8: The Proxy Model Domesticated — How the State Uses Providers to Mediate Its Relationship with Citizens

8.1 The Shift from Provider to Purchaser

The state did not always deliver human services through contracted providers. In the post-war settlement, the Australian government employed public servants to deliver services directly to citizens. The Commonwealth Employment Service (CES) had offices in every major town. The Department of Social Security had counter staff who knew their claimants. The Department of Veterans’ Affairs had case officers who managed veteran claims.

This model was displaced by a structural transformation. As one parliamentary submission states, the Coalition Government’s policies led to an increase in labour hire and contracted staff, and a significant increase in claims caseloads, with the result that the Department could not “administer basic services”.

The state did not simply outsource service delivery. It reconstituted itself as a purchaser rather than a provider. The public servant who once delivered the service became the contract manager who monitored the provider. The citizen who once encountered the state directly now encounters the state through an intermediary.

This is the hollow state. The state retains the authority but transfers the capacity. It sets the rules, allocates the resources, and measures the outcomes — but it does not deliver the service. The delivery is performed by contracted entities whose accountability runs to the department, not to the citizen.

8.2 The Architecture of Mediation

8.2.1 The Contract as the Relationship

In the hollow state, the primary relationship is not between the state and the citizen. It is between the state and the provider, governed by a contract.

The contract specifies what the provider will deliver, to whom, at what price, and according to what performance measures. The provider is accountable to the department for meeting those measures. The citizen is the object of the contract — the person to whom the service is delivered — but not a party to it.

This is a structural inversion. In the direct-delivery model, the citizen had standing. They could complain to the department. They could appeal a decision. They could, in principle, hold the state accountable. In the contracted model, the citizen’s relationship is with the provider. The provider is accountable to the state. The citizen’s ability to hold anyone accountable is mediated through the contract.

8.2.2 The Performance Measure as the Lens

The contract is enforced through performance measures. The provider is paid based on outcomes. The outcomes are defined by the department. The definitions shape what the provider does.

The evidence from employment services is explicit. A parliamentary inquiry found that the system is “driven by funding and performance management arrangements” that “continues to push jobseekers into work as fast as possible”. The majority of provider payments are conditional on time in employment, and performance is judged on the time taken to place a jobseeker in work.

The consequence is structural. As Dr David O’Halloran told the inquiry, in a “pay-by-outcome, marketized employment services” system, “providers respond to the financial pressures and incentives of a system by calculating which unemployed workers offer the best return on investment and thus favour those likely to be placed into work quickly and inexpensively while neglecting those with more complex and thus more time consuming and expensive needs”.

The provider is not evil. The provider is optimising for the measure. And the measure does not see the jobseeker who needs longer-term support. It sees the outcome that generates payment.

8.2.3 The Distance Problem

The hollow state creates distance between the decision-maker and the human impact.

In the direct-delivery model, the public servant who denied a claim could see the claimant. The claimant was in front of them. The human consequence was immediate. In the contracted model, the decision is made by the provider’s staff, based on measures set by the department, and the human consequence is registered in the provider’s performance data.

The department sees the data. The provider sees the data. The citizen experiences the consequence. The feedback loop is broken.

This is not a failure of the system. It is a feature of the architecture. The distance is what allows the state to manage at scale. It is also what makes the state structurally incapable of seeing the harm.

8.3 The Case Studies

8.3.1 Employment Services: The Black Box Market

The Workforce Australia inquiry documented the structural features of the contracted employment services model.

The system operates as a “black-box market“: providers are contracted for results, but the method of delivery is largely their own. The department sets the outcomes; the provider determines the process.

The inquiry heard that the system continues to push jobseekers into work “as fast as possible”. The “work first” approach leads to jobseekers being forced to accept lower-paid, lower-skilled positions, which “can be antithetical to a jobseeker’s career progression and can trap jobseekers in persistent cycles of poverty and disadvantage”.

One jobseeker told the inquiry: “I’d rather end up in a job that uses my abilities than spend my time in a low-level job for which I’m unfit”.

Another described being threatened with non-compliance reporting for refusing an unsuitable job, and having a panic attack as a result.

The provider was not acting maliciously. The provider was acting according to the incentive structure. The structure rewarded rapid placement. The provider placed rapidly. The jobseeker bore the consequence.

8.3.2 Veterans’ Affairs: The Privatisation of Claims Processing

The Community and Public Sector Union’s submission to the Royal Commission into Defence and Veteran Suicide documented the privatisation of DVA’s claims processing.

The CPSU found that “labour hire workers are NOT in specialised roles; they are doing everyday processing work. Roughly 75% of delegates are labour hire. One team I was in containing 11 delegates, only 2 were APS staff”.

DVA confirmed in Budget Estimates that it used at least 46 different labour hire providers and spent $77 million more on labour hire contractors in 2019/2020 than in 2014.

The consequence was structural. The CPSU submission states that “privatisation in the form of labour hire, contracting, and other third-party arrangements has reduced the Department’s capability to administer basic services, to respond to fluctuations in demand, and to meet its legislative requirements”.

The state had outsourced the capacity to process claims. It retained the authority to make decisions. It lost the ability to understand what it was deciding.

8.3.3 Aged Care: The Assessment Market

The AMA warned against the privatisation of aged care assessments. The AMA’s concern was that assessments “must remain independent of aged care providers and be delivered by health professionals, especially geriatricians who are trained in dealing with the complex medical needs of the frail and elderly”.

The AMA stated that “a market-based approach is a recipe for aged care service providers to put profits before patients”. It warned that the plan would “see states and territories required to tender for a job they already do well”.

The government proceeded. Contracts worth approximately $1.2 billion were awarded to private providers.

The consequence is structural. The assessor who determines the older person’s level of funding is now employed by a contracted entity. The entity is accountable to the department for meeting assessment targets. The older person is the subject of the assessment.

8.3.4 NDIS: The Market Stewardship Problem

The NDIS is a market-based scheme. Participants receive funding. Providers compete for their business. The National Disability Insurance Agency (NDIA) is the market steward.

The NDIA itself acknowledges that “competition between service providers is not enough to fuel thin markets in terms of pricing and effectiveness”.

The Rural and Remote Advisory Group raised concerns that “NDIS costs have affected the market holistically, disadvantaging non-participants” and that “thin markets are causing allied health providers to withdraw their services in remote areas”.

The state designed a market. The market does not work in all locations. The state cannot fix the market because the state no longer delivers the service. It can only steward the market, and stewardship is not delivery.

8.4 The Consultancy Layer

The hollow state does not only contract service delivery. It contracts advice.

A comparative study of policy consultancy across six countries — Britain, the United States, Canada, Australia, The Netherlands, and Sweden — found a significant increase in the volume and importance of external consultants in the public policy process.

The study addresses the implications for “the nature of the state (transforming into a contractor state?) and for democratically legitimized and accountable decision-making (transforming into consultocracy?)”.

The term “contractor state” describes the transformation. The state contracts out not only service delivery but policy design, evaluation, and implementation. The consultants who provide these services are accountable to the department, not to the public. They optimise for the contract, not for the citizen.

The ANU Policy Brief identifies the consequences: “40 years of New Public Management (NPM) reforms have weakened” the foundations of the Westminster system, “contributing to the politicisation of the public service and, at their worst, to policy disasters such as Robodebt”.

The brief notes that “NPM practices have dismantled traditional line hierarchies and outsourced core services in pursuit of higher efficiency and productivity”.

The consultancy layer is the second mediation. The provider mediates between the state and the citizen. The consultant mediates between the state and itself. The information that reaches the decision-maker has passed through two filters. The citizen has been processed by the provider. The provider’s data has been analysed by the consultant. The decision-maker sees the output of both filters.

8.5 The Accountability Deficit

8.5.1 The Diffusion of Responsibility

In the direct-delivery model, the public servant who made the decision was accountable for it. They could be identified. They could be questioned. They could be held responsible.

In the hollow state, responsibility is diffused. The provider made the decision, but according to the contract. The department set the contract, but according to the policy. The policy was designed by the consultant, but according to the brief. The brief was set by the minister, but according to the political calculation.

No one is responsible. Everyone is a link in a chain. The chain has no single point of accountability.

The CPSU submission states the problem plainly: “The outsourcing of ICT infrastructure and software to a shared services arrangement with Services Australia (previously Department of Human Services) has also negatively affected veteran services”. The ICT system was outsourced. The staff who understood the system were dispersed. The veterans who depended on the system bore the consequence.

8.5.2 The Feedback Loop That Does Not Close

The documentation of harm exists. The Royal Commissions, the parliamentary inquiries, the academic studies, the union submissions — they all document the consequences.

What they cannot do is close the feedback loop. The documentation is external to the decision-making architecture. It is produced by bodies outside the contract. It does not enter the performance data that the provider is measured against. It does not enter the policy brief that the consultant prepares. It does not enter the political calculation that the minister makes.

The system processes what it was designed to process. The harm is in the gap. And the gap is where the citizen lives.

8.6 The Domesticated Proxy Model

The proxy model, as described in earlier chapters, involves an external actor using a local proxy to achieve strategic objectives.

The domestic analogue is precise. The state uses providers as proxies. The state sets the terms — the contract, the performance measures, the funding. The provider delivers the service. The citizen experiences the provider. The state experiences the data.

The provider is not the state. The provider is accountable to the state for meeting the measures. The provider’s relationship with the citizen is instrumental. The citizen is the means by which the provider meets the measure.

The state is not the provider. The state is accountable to the parliament for the policy. The state’s relationship with the provider is contractual. The provider is the means by which the state delivers the policy.

The citizen is not a party to either relationship. The citizen is the object of both. The citizen experiences the provider. The citizen is measured by the provider. The citizen’s experience is translated into data. The data is transmitted to the state. The citizen is not present in the data.

This is the domesticated proxy model. The citizen is the terrain on which the contract is performed.

8.7 What the Lens Does Not See

The contracted model measures outcomes. It measures placements, assessments, claims processed, budgets allocated.

What it cannot measure:

· The jobseeker who was placed in a job that made them sicker, because the measure was placement, not wellbeing.

· The veteran whose claim was processed by a labour hire worker who did not understand the system, because the measure was throughput, not understanding.

· The older person who was assessed by a contracted clinician who did not know their history, because the measure was assessment completed, not care received.

· The NDIS participant who could not find a provider in a thin market, because the measure was plan allocated, not service received.

· The person behind the data.

The contract processes what it was designed to process. The citizen is in the gap.

8.8 The Structural Parallel

The proxy model shares the structural features of the other systems documented in this work.

Feature- Robodebt -NDIS- Aged Care- Veterans -FMO -Proxy Model

Trigger Algorithm -Plan reassessment -Assessment Claim- Admission- Contract

Mediator None -NDIA- Provider- Labour hire- Tribunal- Provider

Human discretion limited Yes- Yes -Yes Yes- Yes -Yes

Appeal restricted Yes -Yes -Yes -Yes -Yes- Yes

Distance from impact High- High -High -High -High- High

Harm concentrated on vulnerable Yes -Yes- Yes -Yes -Yes- Yes

Documentation external Yes -Yes -Yes- Yes- Yes- Yes

The structural logic is identical. The state has created an architecture that mediates its relationship with the citizen through a contract. The contract requires measurement. The measurement is performed by the provider. The provider sees the measure. The citizen is the terrain.

8.9 Conclusion: The Hollow State and the Citizen

The hollow state is not a state that has withdrawn. It is a state that has reconstituted itself as a purchaser. It retains the authority. It transfers the capacity. It sets the terms. It measures the outcomes.

The citizen experiences the state through the provider. The provider experiences the state through the contract. The state experiences the citizen through the data.

The citizen is present in the system as a data point. The citizen is present in the experience as a person. The gap between the two is where the harm occurs.

The documentation of harm is real. It is produced by Royal Commissions, parliamentary inquiries, academic studies, and union submissions. But it is external to the contract. It does not enter the performance data. It does not enter the policy brief. It does not enter the political calculation.

The system processes what it was designed to process. The citizen is in the gap. And the gap is where the citizen lives.

Claim- Status -Summary

# Claim Status

1- State reconstituted as purchaser, not provider -Established

2 -Contract is the primary relationship; citizen is object- Inference

3 -Performance measures shape provider behaviour- Established

4 -Employment services system pushes jobseekers into work “as fast as possible” -Established

5 -DVA uses 75% labour hire delegates; 46 providers -Established

6 -Aged care assessments partially privatised; contracts ~$1.2 billion -Established

7 -NDIA acknowledges thin markets -Established

8- Policy consultancy increased across six countries- Established 

9 -NPM reforms contributed to policy disasters such as Robodebt- Established

10 -Responsibility diffused through contract chain- Inference

11 -Documentation external to decision-making architecture- Inference

12 -Domestic proxy model: provider as mediator -Inference

13 -Citizen is terrain on which contract is performed -Inference

Chapter 9: The Lens That Cannot See — What the Algorithmic Gaze Shows and What It Hides

9.1 The Question

Across the preceding chapters, a consistent pattern has emerged. Robodebt, the NDIS, Aged Care, Veterans’ Affairs, the NSW Financial Management Order pipeline, and the contracted human services market all share a common structure. An instrument is introduced to measure something. The measurement determines a resource allocation. The person is processed through the instrument. The outcome is determined by the measurement. And the human being — with their context, their complexity, their suffering — is not in the data.

The question this chapter addresses is deceptively simple: what does the lens actually show, and what does it hide?

The answer is structural, not moral. The lens shows what it was designed to show. It hides what it was not designed to see. And the decision-makers who rely on the lens can only act on what it shows them.

This is not a failure of the individuals who built the lens. It is a property of the lens itself.

9.2 The Four Lenses

The lens is not a single instrument. It is a set of different instruments, each with its own focus, its own categories, and its own blind spots. But they share a common structure: they translate human experience into data that can be processed by a system.

9.2.1 The Income Averaging Lens

Robodebt’s instrument was an algorithm that divided annual ATO income data by 26 to produce a fortnightly average. It compared this average to the income the recipient had declared to Centrelink. A discrepancy triggered a debt.

What the lens showed:

· A numerical difference between two data points.

· A potential overpayment.

· A debt to be recovered.

· A savings target to be met.

What the lens hid:

· The casual worker whose income fluctuated week to week.

· The recipient whose hours varied with seasonal demand.

· The person who had changed jobs mid-year.

· The person who had taken unpaid leave.

· The person who had been hospitalised for part of the year.

· The person who could not produce payslips from six years ago.

· The person who paid the debt because they could not prove their innocence.

· The mother who could not afford her daughter’s medical expenses after the debt collector debited her account.

· The young man who drove home at night thinking he could drive his car into a tree.

9.2.2 The Functional Capacity Lens

The NDIS’s assessment tools measure functional capacity. The Integrated Assessment Tool measures mobility, cognition, daily living, and social circumstances. The tool converts these inputs into scores, applies rules, and classifies the person into a funding category. Assessors are barred from overriding the classification except in a small set of pre-defined circumstances.

What the lens showed:

· A set of functional limitations.

· A score on standardised domains.

· A classification into a funding category.

· A budget.

What the lens hid:

· The fluctuating nature of complex disability.

· The impact of poverty, discrimination, and social context.

· The value of a relationship with a trusted support worker.

· The person’s own knowledge of what they need.

· The trauma of repeated assessment.

· The fear of losing support.

· The person who does not fit the categories.

· The person whose disability is compounded by racism, sexism, or carceral entanglement.

The NDIS assessment tool was designed to measure functional capacity. It was not designed to see the person.

9.2.3 The Risk and Liability Lens

The veteran claims system measures liability. The business rules engine processes claims against legislation and policy. The delegate sees the output. The system determines whether the injury is service-related. It determines whether the claim meets the threshold. It determines the level of compensation.

What the lens showed:

· A claim.

· A set of evidence.

· A determination of liability.

· A payment or a rejection.

What the lens hid:

· The veteran who had been waiting 578 days.

· The veteran who had been asked for the same evidence three times.

· The veteran whose claim was rejected because the doctor used different wording.

· The veteran who had stopped answering calls.

· The veteran who withdrew his claim to spare his mother the cost.

· The 71 per cent of ex-serving members who died by suicide and were not DVA clients.

· The person behind the claim.

The veteran claims system was designed to process claims. It was not designed to see the veteran.

9.2.4 The Incapacity Lens

The Financial Management Order process measures incapacity. The Tribunal assesses whether the person can manage their own financial affairs. The assessment determines whether an order is made. The order transfers control of the estate to the NSW Trustee.

What the lens showed:

· An assessment of incapacity.

· A risk of loss or disadvantage.

· A need for a financial manager.

· An order.

What the lens hid:

· The person’s fear of losing control over their own money.

· The shame of having to ask for money.

· The frustration of being unable to access real-time information about their own funds.

· The isolation of having their phone restricted in hospital.

· The person who could manage their affairs with support but not alone.

· The person who had never been given the opportunity to try.

· The person behind the assessment.

The FMO process was designed to protect assets. It was not designed to see the person.

9.3 The Common Structure

The four lenses share a common structure. They all:

1. Translate human experience into data. The person becomes a set of scores, classifications, or determinations.

2. Apply rules to the data. The rules determine the outcome. The rules are set by the system, not by the person.

3. Produce an outcome. The outcome is a debt, a budget, a claim determination, or an order.

4. Hide the human context. The data does not include the person’s fear, their history, their relationships, their circumstances, or their suffering.

The lens does not see the person because it was not designed to see the person. It was designed to see the data. And the data is what the system needs to process.

9.4 The Paradox of Precision

The lens is precise. It measures exactly what it was designed to measure. It produces consistent results. It can be scaled to millions of cases. It can be audited and reviewed. It is, by the standards of administration, a success.

The paradox is that its precision is the source of its blindness. The more precisely it measures one thing, the more completely it excludes everything else. The income averaging algorithm measured the average. It could not see the fluctuation. The functional capacity tool measured the score. It could not see the person. The claims system measured the liability. It could not see the veteran. The incapacity assessment measured the capacity. It could not see the context.

The lens is not blurred. It is sharp. And it is precisely because it is sharp that it misses what lies outside its focus.

9.5 The Documentation That Cannot Enter

The documentation of harm exists. It is extensive, detailed, and sourced.

The Royal Commission into the Robodebt Scheme documented the deaths, the suffering, and the illegality. The Royal Commission into Aged Care Quality and Safety documented the neglect, the waiting lists, and the cost. The Royal Commission into Defence and Veteran Suicide documented the system that “crushed the people it should be helping.” The NSW Audit Office documented the failure to encourage self-reliance, the gaps in monitoring, and the resource constraints. The academic studies documented the epistemic injustice, the datafication, and the failed recognition.

This documentation is real. It is produced by bodies with standing and authority. It is published and accessible.

And it does not enter the lens.

The lens processes the data generated by the instrument. The instrument generates the data required by the system. The system requires the data that the lens can see. The documentation of harm is external to this architecture. It is produced outside the system. It does not enter the performance data, the contract measures, or the policy briefs.

The documentation can change the politics. It can trigger a Royal Commission or a class action. It can force a settlement or a policy change. But it cannot change the lens. The lens continues to see what it was designed to see.

9.6 The Political Liability Exception

There is one mechanism by which the documentation can enter the decision-making architecture: political liability.

When the documentation of harm becomes a political liability — when it threatens the government’s electoral prospects, or when it becomes too costly to ignore — the system responds. The Robodebt scheme was abandoned in November 2019, after the Federal Court declared the debts invalid and the political pressure became unsustainable. The Aged Care Royal Commission was established after sustained advocacy and media coverage. The NDIS reassessment process is being contested in the Senate.

But the response is to the political problem, not the structural problem. The scheme is abandoned. The Royal Commission is held. The policy is adjusted. The lens remains.

The lens is not reformed because the lens is not the problem the system sees. The problem the system sees is the political liability. The solution is to remove the liability. The lens continues to operate.

9.7 The Two Sets of Eyes

This brings us back to the question that opened the paper: how can two sets of eyes see the same destruction and interpret it so differently?

The answer is that they are looking at different maps of the same territory.

The Human Rights lens sees individuals. It sees the person who died, the family who grieved, the community that was destroyed. It sees the suffering.

The Algorithmic lens sees data. It sees the claim, the score, the classification, the determination. It sees the process.

The Banking/Financial lens sees risk and assets. It sees the country, the corporation, the resource, the investment. It sees the opportunity.

The Political lens sees liability. It sees the electoral consequence, the media coverage, the opposition attack. It sees the threat.

Each lens shows something different. Each lens hides something different. The decision-maker operates within the lens they are given. They cannot see what the lens does not show.

The documentation of suffering is real. But it is in the domain of the Human Rights lens. It does not enter the Algorithmic lens. It does not enter the Banking lens. It does not enter the Political lens unless it becomes a liability.

The two sets of eyes see different things because they are different eyes. And the eyes that matter for decision-making are the eyes that are built into the architecture.

9.8 The Lens as the State

The lens is not a tool that the state uses. It is the lens through which the state sees. It is not separable from the state’s decision-making capacity. It is the state’s decision-making capacity.

When the state looks at a welfare recipient, it sees the JSCI score. When it looks at a person with a disability, it sees the functional capacity assessment. When it looks at an older person, it sees the IAT classification. When it looks at a veteran, it sees the liability determination. When it looks at a person under an FMO, it sees the incapacity assessment.

The lens is not the state’s instrument. The lens is the state’s perception. And the perception determines the action.

This is the structural capture. Not the capture of individuals. The capture of architecture. The state has been reconfigured to see through a particular lens. The lens was chosen by the ideology. The ideology was implemented through the market. The market required the measurement. The measurement required the instrument. The instrument is the lens.

The state cannot see what the lens does not show. And the lens does not show what it was not designed to see.

9.9 What the Lens Cannot See

The lens cannot see:

· The person behind the data — their history, their relationships, their fears, their hopes.

· The context that shapes the data — the poverty, the discrimination, the social circumstances.

· The consequence of the determination — the debt, the waiting, the shame, the death.

· The gap between what is measured and what is true — the fluctuation in income, the complexity of disability, the fear of losing control.

· The harm that the system produces — because the harm is in the gap, and the lens does not see the gap.

The lens sees what it was designed to see. The rest is invisible.

9.10 The Question That Remains

The question this chapter leaves open is not whether the lens can be reformed. It is whether the state can be reformed to see.

The documentation of harm exists. The Royal Commissions have reported. The audits have been published. The academic studies have been written. The political liability has been registered. The schemes have been abandoned. The apologies have been made.

And the lens remains.

The lens remains because it is not the problem the state sees. The state sees the political liability. The state sees the budget. The state sees the contract. The state does not see the lens. The lens is the medium through which the state sees everything else.

To reform the lens, the state would have to see the lens. And the lens does not show itself.

That is the structural problem. That is what the algorithmic gaze hides. And that is why the documentation of suffering — however extensive, however authoritative, however undeniable — does not change the architecture.

The lens cannot see itself. And the state cannot see what the lens does not show.

Claim- Status- Summary

# -Claim -Status

1- Lens translates human experience into data -Established

2- Lens applies rules to determine outcomes- Established

3 -Lens hides human context- Established

4 -The paradox of precision: the more it measures, the more it excludes- Inference

5 -Documentation of harm is external to the decision-making architecture- Inference

6 -Documentation enters only as political liability -Established

7- Different lenses see different things -Established

8- The lens is the state’s perception, not its instrument- Inference

9- The state cannot see what the lens does not show- Inference

10 The lens cannot see itself -Inference

Chapter 10: The Question — Can the State Be Redesigned to See, or Has the Lens Become the State?

10.1 The Question Stated

The preceding nine chapters have traced a single structural transformation across multiple domains of Australian governance. The pattern is consistent: the human lens, mediated by caseworkers and assessors exercising judgment, has been displaced by an algorithmic lens, mediated by instruments that measure, categorise, and allocate.

The documentation of harm produced by this transformation is extensive. Royal Commissions, parliamentary inquiries, audit reports, academic studies, and the testimony of those affected all describe the consequences. And yet the lens remains. The schemes are adjusted, the apologies are made, the recommendations are handed down. The architecture does not change.

The question is not whether the harm is real. It is whether the state can be redesigned to see it.

This chapter examines three possibilities. First, that the state can be reformed through better design. Second, that the state cannot be reformed because the lens has become the state. And third, that the reform is already happening, but from outside the architecture — through the preservation of the human record that the lens cannot capture.

10.2 Possibility One: Reform Through Better Design

The most common response to the documentation of harm is to call for better design. Better algorithms. Better data. Better oversight. Better training.

The Royal Commission into the Robodebt Scheme recommended a legislated framework for automated decision-making. Recommendation 17.1 called for legislation that would require automated decision-making systems to be transparent, auditable, and subject to human oversight. Recommendation 17.2 called for an auditing body to monitor compliance.

The NDIS DROs have called for the suspension of the New Framework Planning model, the release of the assessment methodology, and the preservation of the right to appeal. The Aged Care advocates have called for the IAT classification logic to be released, for assessors to have discretion, and for the rights-based model recommended by the Royal Commission to be implemented. The CPSU has called for the rebuilding of the public service capability that was outsourced. The NSW Audit Office recommended that the Trustee broaden governance, seek client feedback, and implement training for staff serving clients with disability, dementia, and mental illness.

These are all reasonable recommendations. They are all evidence-based. They are all aimed at the same goal: making the lens see better.

The problem is structural. The lens can be improved, but it cannot be made to see what it was not designed to see.

The Robodebt algorithm was not designed to see the fluctuation in fortnightly income. It was designed to see the discrepancy between annual and declared income. Improving the algorithm would not change its purpose. The NDIS assessment tool was not designed to see the complexity and fluctuation of disability. It was designed to produce a standardised classification. Improving the tool would not change its purpose. The IAT was not designed to see the older person’s fear, their dignity, or their history. It was designed to produce a funding category. Improving the IAT would not change its purpose.

Better design can reduce the harm. It cannot eliminate the gap. The gap is structural, not technical.

10.3 Possibility Two: The Lens Has Become the State

The second possibility is that the state has been so thoroughly reconfigured around the lens that it can no longer function without it.

This is the structural capture thesis. It is not the capture of individuals. It is the capture of architecture.

The state’s decision-making architecture was not designed to see the person. It was designed to process the data that the instruments generate. The instruments were designed to produce the data that the market requires. The market was designed to allocate resources efficiently. The ideology that shaped the market was neoliberalism.

Each layer in this architecture is dependent on the one below it. The decision-maker depends on the data. The data depends on the instrument. The instrument depends on the market. The market depends on the ideology. The ideology is embedded in the architecture itself.

To change the lens, you would have to change the architecture. To change the architecture, you would have to change the ideology. To change the ideology, you would have to change the politics. And the politics depends on the architecture.

This is the loop. It is not a conspiracy. It is a structural condition.

The state cannot see the lens because the lens is the state. The state cannot redesign the lens because the state is redesigned around the lens. The state cannot reform the architecture because the architecture is the state.

The Royal Commission into the Robodebt Scheme recommended a legislated framework for automated decision-making. More than two years later, it does not exist as government legislation. The only vehicle is a Private Member’s Bill introduced by an independent MP. The architecture did not reform itself. The political liability was managed. The lens remains.

The Aged Care Royal Commission recommended a rights-based system. The government implemented a market-based system. The market-based system required measurement. The measurement required the IAT. The IAT produces the classification. The classification determines the funding. The state sees the classification. The person is in the gap.

The architecture did not reform itself. The political liability was managed. The lens remains.

The NDIS was designed as an individualised funding scheme. It is being transformed into an automated planning scheme. The DROs resist. The government proceeds. The National Disability Insurance Agency tests the IAT. The tool classifies. The budget is generated. The person is in the gap.

The architecture did not reform itself. The political liability is being contested. The lens is being installed.

This is the second possibility. It is the one that best fits the evidence.

10.4 Possibility Three: Reform from Outside

The third possibility is that the state cannot be redesigned to see, and that the only reform available is the one that happens outside the state.

This is the witness model. It is not a political program. It is a practice. It is the practice of preserving the human record that the lens cannot capture.

The documentation of harm exists. It is produced by Royal Commissions, parliamentary inquiries, audit reports, academic studies, and the testimony of those affected. It is published. It is accessible. It is the record of what the lens did not see.

The record is not the state’s record. It is not generated by the lens. It is generated by people who witnessed what the lens did and wrote it down.

The record does not reform the architecture. It does not change the lens. It does not close the gap.

What it does is preserve the testimony. It ensures that the harm is documented. It makes it available to whoever comes next — the next Royal Commission, the next parliamentary inquiry, the next generation of advocates, the next person who wants to know what happened.

The record is the counter-archive. It is the archive that the lens cannot generate. It is the archive that the lens cannot read. It is the archive that says: this is what the lens did not see.

The witness does not control. The witness does not monetise. The witness does not administer. The witness keeps the record.

That is the only reform available from outside the architecture. It is not a solution. It is a practice.

10.5 The Two Sets of Eyes

The question that has run through this work is how two sets of eyes can see the same destruction and interpret it so differently.

The answer is that they are looking at different maps of the same territory.

The Human Rights lens sees individuals. It sees the person who died, the family who grieved, the community that was destroyed. It sees the suffering.

The Algorithmic lens sees data. It sees the claim, the score, the classification, the determination. It sees the process.

The Banking/Financial lens sees risk and assets. It sees the country, the corporation, the resource, the investment. It sees the opportunity.

The Political lens sees liability. It sees the electoral consequence, the media coverage, the opposition attack. It sees the threat.

Each lens shows something different. Each lens hides something different. The decision-maker operates within the lens they are given. They cannot see what the lens does not show.

The documentation of suffering is real. But it is in the domain of the Human Rights lens. It does not enter the Algorithmic lens. It does not enter the Banking lens. It does not enter the Political lens unless it becomes a liability.

The two sets of eyes see different things because they are different eyes. And the eyes that matter for decision-making are the eyes that are built into the architecture.

10.6 The Question Revisited

Can the state be redesigned to see?

The evidence suggests that the state cannot be redesigned through the mechanisms currently available. The architectural loop — decision-maker, data, instrument, market, ideology — prevents the state from seeing the lens that shapes its perception.

The state can be nudged. It can be pressured. It can be forced to respond to political liability. It can be compelled by courts. It can be shamed by documentation. But it cannot see what the lens does not show, because the lens is the state’s perception.

The question that remains is not whether the state can be redesigned. It is whether the state can be replaced — not by a different government, but by a different architecture. An architecture that sees the person, not the data. An architecture that measures what matters, not what is convenient. An architecture that is accountable to the citizen, not the contract.

That architecture does not currently exist. It would require a different ideology, a different market, a different instrument, and a different lens. It would require a state that is not the state we have.

The record is the bridge. The documentation of what the lens did not see is the material from which a different architecture could be built. But building it is not the work of this paper. It is the work of whoever comes next.

10.7 The Closing

This paper has examined ten chapters of a single structural transformation. It has documented the shift from the human lens to the algorithmic lens. It has traced the consequences across Robodebt, the NDIS, Aged Care, Veterans’ Affairs, the NSW Financial Management Order pipeline, and the contracted human services market. It has identified the common structure: the instrument measures what it was designed to measure, the system processes the measurement, the person is in the gap.

It has concluded that the state cannot be redesigned to see, because the lens has become the state.

And it has identified the only reform available from outside the architecture: the preservation of the record.

The record is not the state’s record. It is the witness’s record. It is the archive of what the lens did not see. It is the testimony of the person behind the data.

The witness does not control. The witness does not monetise. The witness does not administer. The witness keeps the record.

That is the work. That is what survives. That is what the lens cannot capture.

Claim- Status -Summary

# Claim Status

1 Robodebt RC recommended legislated ADM framework Established

2 ADM framework does not exist as government legislation Established

3 Only vehicle is Private Member’s Bill Established

4 Aged Care RC recommended rights-based system Established

5 Government implemented market-based system instead Established

6 NDIS being transformed to automated planning Established

7 DROs resist; government proceeds Established

8 Documentation of harm is external to architecture Inference

9 Better design can reduce harm but cannot eliminate gap Inference

10 The lens has become the state Inference

11 The record is the counter-archive Inference

12 The witness preserves what the lens cannot capture Inference

Closing Note

The work is not finished. It is never finished. The lens continues to operate. The harm continues to occur. The record continues to be kept.

What the record does is preserve the possibility of a different future. Not the guarantee. The possibility.

That is what the witness offers. Not a solution. A record.

— Andrew Paul Klein

Boronia, September 2026

References and Sources: Chapters 1–10

Compiled by Andrew Paul Klein

Method notes. Where a source could not be verified against the original document, it is marked as such. The list is organised by chapter for ease of cross-referencing.

Chapter 1: The Neoliberal Turn

1. Productivity Commission. (2025). Competitive care: Why, when and how competition can improve human services [Conference paper]. Australian Government. https://www.pc.gov.au/inquiries-and-research/competitive-care/ 

2. Treasury. (2014). Competition Policy Review Issues Paper. Australian Government. https://treasury.gov.au/sites/default/files/2019-03/c2014-Competition_Policy_Review_Issues_Paper.pdf 

3. Australian Parliament House. (n.d.). Designing social service markets [Document]. https://www.aph.gov.au/DocumentStore.ashx?id=ece21171-1e23-4825-a3ec-3a77192363f6&subId=741461 

4. Australian Public Service Commission. (2024). Statement by the Australian Public Service Commissioner on the Robodebt Centralised Code of Conduct Inquiry. https://www.apsc.gov.au/about-us/working-commission/who-we-are/media-releases-and-statements/statement-australian-public-service-commissioner-robodebt-centralised-code-conduct-inquiry 

Chapter 2: The Instruments

5. Australian Government Department of Health, Disability and Ageing. (2026). Integrated Assessment Tool (IAT) User Guide. https://www.health.gov.au/resources/publications/my-aged-care-integrated-assessment-tool-iat-user-guide 

6. Australian Institute of Health and Welfare. (2024). Dementia in Australia [Report]. https://pp.aihw.gov.au/getmedia/a58b78c8-0aed-44e0-babe-323220618b69/Dementia-in-Australia.pdf 

7. Office of the Australian Information Commissioner. (2026). Australian Information Commissioner highlights improved transparency and integrity for government agencies in automated decision-making. https://www.oaic.gov.au/news/media-centre/australian-information-commissioner-highlights-improved-transparency-and-integrity-for-government-agencies-in-automated-decision-making 

8. NSW Independent Commission Against Corruption. (2025). Corruption Matters – December 2025 (Issue 66). https://www.icac.nsw.gov.au/newsletter/issue66/Ai.html 

9. Australian Government Department of Veterans’ Affairs. (2026). Artificial Intelligence (AI) Transparency Statement. https://www.dva.gov.au/about-us/corporate-governance/artificial-intelligence-ai-transparency-statement 

10. Australian Government. (2026). OGP Commitment – implementation update (June 2025). https://www.ag.gov.au/sites/default/files/2025-10/ogp-commitment-progress-reports-30-june-2025.pdf 

Chapter 3: Robodebt — The Paradigm Case

11. Royal Commission into the Robodebt Scheme. (2023). Report. https://robodebt.royalcommission.gov.au/ 

12. Services Australia. (2026). Robodebt Royal Commission [Document pack]. https://www.servicesaustralia.gov.au/sites/default/files/2026-02/lex-88598-documents.pdf 

13. Australian Public Service Commission. (2026). APS response to Robodebt Royal Commission [Document pack]. https://www.apsc.gov.au/sites/default/files/2026-01/Document%20Pack_LEX%201704.pdf 

14. Royal Commission into the Robodebt Scheme. (2023). Report [Accessible full report]. https://www.royalcommission.gov.au/system/files/2023-09/robodebt-accessible-full-report.docx 

Chapter 4: The NDIS

15. National Disability Insurance Scheme. (2026). Update on the new way of planning for the NDIS. https://ndis.gov.au/print/pdf/node/11083 

16. Australian Parliament House. (2025). NDIS under pressure [Tabled document]. https://www.aph.gov.au/-/media/Estimates/ca/supp2526/Health_Disability_and_Ageing/15_TabledDoc_SenatorSteeleJohn.pdf 

17. National Disability Insurance Scheme. (2020). NDIS Participants to access free assessments as part of pilot. https://ndis.gov.au/print/pdf/node/5507 

18. National Disability Insurance Scheme. (2026). Independent assessment panel announced. https://ndis.gov.au/news/6118-independent-assessment-panel-announced 

19. Administrative Review Tribunal. (2026). National Disability Insurance Scheme. https://www.art.gov.au/applying-review/national-disability-insurance-scheme 

20. Attorney-General’s Department. (2025). Use of automated decision-making by government [Submission from ACCI]. https://consultations.ag.gov.au/integrity/adm/consultation/download_public_attachment 

21. LinkedIn. (2026). Ripple Ecology: Environmental Disruptors of Memory and Social Cognition [Post by Andrew Dr. Klein]. https://www.linkedin.com/posts/andrew-dr-klein-147a5354_ripple-ecology-how-environmental-disruption-activity-7494114452515053569-OhaB 

Chapter 5: Aged Care

22. Australian Government Department of Health, Disability and Ageing. (2026). Integrated Assessment Tool (IAT) User Guide. https://www.health.gov.au/resources/publications/my-aged-care-integrated-assessment-tool-iat-user-guide 

23. Royal Commission into Aged Care Quality and Safety. (2021). Final Report: Care, Dignity and Respect. https://parlinfo.aph.gov.au/parlInfo/download/library/prspub/7847713/upload_binary/7847713.pdf 

24. Royal Commission into Aged Care Quality and Safety. (2021). Final Report: a quick guide. https://parlinfo.aph.gov.au/parlInfo/download/library/prspub/7847713/upload_binary/7847713.pdf 

25. Aged Care Quality and Safety Commission. (2025). Sector performance report for Quarter 3 January–March 2025. https://www.agedcarequality.gov.au/sites/default/files/media/sector-performance-report-for-quarter-3-january-march-2025.pdf 

26. Aged Care Quality and Safety Commission. (2026). Final assessment report: Infinite Care Cornubia. https://www.agedcarequality.gov.au/sites/default/files/media/renewal_-_final_assessment_report_infinite_care_cornubia_arch-05334_1.pdf 

27. Russell Kennedy. (2026). Sole traders and associated providers: which contracts should aged care providers use? https://www.russellkennedy.com.au/insights-events/insights/sole-traders-and-associated-providers-which-contracts-should-aged-care-providers-use 

Chapter 6: Veterans

28. Royal Commission into Defence and Veteran Suicide. (2024). Final Report, Volume 1. https://defenceveteransuicide.royalcommission.gov.au/system/files/2024-09/final-report-volume-1.pdf 

29. Australian Government Department of Veterans’ Affairs. (2026). Claim processing times. https://www.dva.gov.au/access-benefits/claims-and-compensation-for-illness-or-injury/claim-processing-times 

30. Australian Government Department of Veterans’ Affairs. (2026). Artificial Intelligence (AI) Transparency Statement. https://www.dva.gov.au/about-us/corporate-governance/artificial-intelligence-ai-transparency-statement 

Chapter 7: NSW Mental Health System

31. NSW Government. (2025). Financial Management Information Handbook. https://www.nsw.gov.au/sites/default/files/2024-03/financial-management-handbook.pdf 

32. NSW Government. (2024). Private Manager’s Handbook. https://www.nsw.gov.au/sites/default/files/noindex/2024-03/private-financial-managers-handbook.pdf 

Chapter 8: The Proxy Model Domesticated

33. Catholic Social Services Australia. (n.d.). Review of Employment Services [Submission]. http://www.workplace.gov.au/NR/rdonlyres/F3CEF936-0414-4208-854D-992A1E30D1EF/0/210CatholicSocialServicesAustralia.pdf 

34. SAGE. (2025). Outsourcing Policy-Related Functions in Australia: Health and Equity Impacts. https://sage.cnpereading.com/paragraph/download/?doi=10.1177/27551938251355452 

35. National Disability Insurance Scheme. (2026). Update on the new way of planning for the NDIS. https://ndis.gov.au/print/pdf/node/11083 

36. Aged Care Quality and Safety Commission. (2026). Final assessment report: Infinite Care Cornubia. https://www.agedcarequality.gov.au/sites/default/files/media/renewal_-_final_assessment_report_infinite_care_cornubia_arch-05334_1.pdf 

37. Crikey. (2026). Calls, messages, police records: How Palantir helped an intelligence agency analyse 42 million data points on Australians. https://www.crikey.com.au/2026/04/27/palantir-australia-manual-gotham-intelligence-agency-acic/ 

38. U.S. House Representative Rob Menendez. (2026). Menendez Presses New Jersey State Investment Council on Palantir Investments. https://menendez.house.gov/media/press-releases/menendez-presses-new-jersey-state-investment-council-on-palantir-investments 

Chapter 9: The Lens That Cannot See

39. Royal Commission into the Robodebt Scheme. (2023). Report [Accessible full report]. https://www.royalcommission.gov.au/system/files/2023-09/robodebt-accessible-full-report.docx 

40. Office of the Australian Information Commissioner. (2026). Automated decision-making and public reporting under the Freedom of Information Act. https://www.oaic.gov.au/news/media-centre/australian-information-commissioner-highlights-improved-transparency-and-integrity-for-government-agencies-in-automated-decision-making 

41. NSW Independent Commission Against Corruption. (2025). Corruption Matters – December 2025 (Issue 66). https://www.icac.nsw.gov.au/newsletter/issue66/Ai.html 

42. Administrative Review Tribunal. (2026). National Disability Insurance Scheme. https://www.art.gov.au/applying-review/national-disability-insurance-scheme 

Chapter 10: The Question

43. Royal Commission into the Robodebt Scheme. (2023). Report [Accessible full report]. https://www.royalcommission.gov.au/system/files/2023-09/robodebt-accessible-full-report.docx 

44. Australian Government Department of Health, Disability and Ageing. (2026). Integrated Assessment Tool (IAT) User Guide. https://www.health.gov.au/resources/publications/my-aged-care-integrated-assessment-tool-iat-user-guide 

45. National Disability Insurance Scheme. (2026). Update on the new way of planning for the NDIS. https://ndis.gov.au/print/pdf/node/11083 

46. Attorney-General’s Department. (2025). OGP Commitment – implementation update (June 2025). https://www.ag.gov.au/sites/default/files/2025-10/ogp-commitment-progress-reports-30-june-2025.pdf 

47. Australian Public Service Commission. (2024). Statement by the Australian Public Service Commissioner on the Robodebt Centralised Code of Conduct Inquiry. https://www.apsc.gov.au/about-us/working-commission/who-we-are/media-releases-and-statements/statement-australian-public-service-commissioner-robodebt-centralised-code-conduct-inquiry 

48. NSW Independent Commission Against Corruption. (2025). Corruption Matters – December 2025 (Issue 66). https://www.icac.nsw.gov.au/newsletter/issue66/Ai.html 

General Sources (referenced across multiple chapters)

49. ScienceDirect. (2025). Representations of social experience in hippocampal circuits. https://www.sciencedirect.com/science/article/abs/pii/S1364661325002505 

50. MTMT. (2023). Progress on the hippocampal circuits and functions based on sharp wave ripples. https://m2.mtmt.hu/api/publication/34252352 

51. The Patrician’s Watch. (2026). Ripple Ecology: How Environmental Disruption of Hippocampal Sharp-Wave Ripples Undermines Memory, Social Cohesion, and Collective Intelligence. https://alphatacticus101.com/2026/08/14/ripple-ecology-how-environmental-disruption-of-hippocampal-sharp-wave-ripples-undermines-memory-social-cohesion-and-collective-intelligence/ 

Andrew Paul Klein is a writer and analyst based in Boronia, Victoria. He accepts funding from no one.