Deep Truth Analysis: The IAT Algorithm Schema May 2026

An institutional reading of the Integrated Assessment Tool, applying Deep Truth Persona Version 5.2 to the official user guide of the Australian Government's aged care algorithm.

Verdict: RedCumulative Moral DisengagementHigh Political Context

Steve Davies | Mindful Progress | @ozloop — Canberra, Australia | 17 May 2026 | Published under Creative Commons

Foreword: What This Analysis Does
The Transparency Gap

The Department of Health and Aged Care has not published the calculation logic that converts an IAT submission into a funding classification. The 411-page User Guide acknowledges "the system" at exactly one point — a single sentence on page 357.

What Deep Truth Does Instead

Deep Truth does not require the algorithm to be published to surface its shape. The user guide leaks the schema through its operational instructions: mandatory inputs, validated tool scores, pre-population flows, eleven discrete outputs, and the override architecture.

The prior question is not whether the algorithm is technically sound — it is whether the public has the information needed to judge that.

Source Inventory

Before mechanism tagging, the source material is inventoried by voice type. The findings are stark.

Institutional / Official Language

Present — primary. The entire 411-page user guide. Anonymous bureaucratic register throughout; no author named on the cover.

Corroborating Witnesses

Absent. No client voices; no assessor voices in their own register appear anywhere in the document.

Critical / Dissenting Voices

Absent. No dissent recorded anywhere in the document. The user guide is monovocal by design.

Documentary Evidence

Partial. References to the Aged Care Act 2024, Aged Care Rules 2025, and MAClearning modules — none reproduced.

Political Context Check
Political Cycle

The Aged Care Royal Commission (2018–2021) found systemic neglect. The Albanese government legislated the Aged Care Act 2024 and launched Support at Home from 1 July 2025. This document operationalises those commitments.

Stakeholder Silence

Providers, unions, COTA, OPAN, IT vendors, Services Australia, and a substantial commercial home care market all have material interests. The user guide names only one stakeholder: the Lead Educators' Network (LEN).

Timing Anomaly

The IAT has been in operation since 1 July 2024 — predating the Aged Care Act 2024 commencement and the Support at Home launch on 1 July 2025. The instrument was operating before its statutory framework was in force.

Step 1 — Contextual Distinction
Voice & Function

Department of Health and Aged Care. Anonymous bureaucratic register throughout. Function: operationalising — instructing the assessment workforce how to enact the algorithmic determination process.

Agency

Distributed and procedurally diffused across: "triage person", "assessor", "clinical assessor", "assessment delegate", "the system", "Services Australia", and "the Department" — roles without consolidated accountability.

How Framing Erases Lives

Institutional language systematically overwrites lived experience:

  • People become "clients"
  • Lives become rated capacities for "transfers", "ability to dress", "handle money"
  • Identity becomes "diversity fields"
  • The Stolen Generations, Forgotten Australians, and Former Child Migrants are collapsed into a single tick box — a workaround because the schema does not recognise care leavers as a category
The Algorithm Schema — Reconstructed

The Department has not published the IAT algorithm. The schema below is inferred from the user guide alone, treating the document as a partial leakage of the system's structure through its instructions to human operators.

The architecture is a queue ordered by an opaque score against a capped pool. Without the weights, no client, advocate, or assessor can determine whether a given case should reasonably fall at a particular classification — and no evidence-based challenge is possible.

Inside the Schema: Inputs & Outputs
~200+ Mandatory Inputs
  • Functional capacity across ~25 ADLs and IADLs (scaled: independent → unable)
  • Validated tool scores: PHQ-4, GP Cog, DEMMI modified, GSGL, Duke SSI, RUIS/RFIS
  • ICD-coded conditions, medication counts, frailty indicators
  • Risk flags: elder abuse, refusal of assistance, harm to self or others
  • Diversity tags: CALD, First Nations, rural/remote, veteran, homeless, LGBTI
  • Carer profile: presence, hours, sustainability
  • Triage priority: High / Medium / Low (assessor-assigned, mandatory)
11 Discrete Output Classifications
  • Ineligible for CHSP/SaH
  • CHSP
  • SaH Classification 1
  • SaH Classification 2
  • SaH Classification 3
  • SaH Classification 4
  • SaH Classification 5
  • SaH Classification 6
  • SaH Classification 7
  • SaH Classification 8

Each maps to a budget envelope under Support at Home. Dollar values are not named in the user guide.

The Override Contradiction

"The system will display a prompt to override to a higher classification. Assessors must not take this action." — IAT User Guide, p. 372

The Architecture

The system generates the IAT outcome as read only. The assessor then ACCEPTs or OVERRIDEs. Overrides are permitted only to: Restorative Care Pathway, End of Life Pathway, Transition Care, Permanent Residential Care, or Residential Respite.

The Contradiction

Where the system displays "Ineligible for CHSP/SaH", it prompts the assessor to override to a higher classification — then the user guide instructs the assessor they must not take this action. The architecture invites the override the policy forbids. The contradiction is not addressed.

Step 2 — Core Analysis: Eight Mechanisms

Reading Guidance under high political context applied. Euphemism is treated as strategic; diffusion is read as politically functional; absence of dissenting voices is a finding.

1
Euphemistic Labelling — Intensity 6

"System generated IAT outcome"; "classification 1–8"; "pre-population". Truthful language: algorithmically determined funding tier within a rationed pool. Renders rationing logic invisible.

2
Displacement of Responsibility — Intensity 6

The Department's policy forbids the override the Department's system invites. Assessors carry moral weight for outcomes they cannot affect; the Department is structurally insulated.

3
Diffusion of Responsibility — Intensity 6

Triage person → assessor → clinical assessor → delegate → "the system" → Aged Care Rules → Department → Services Australia → provider. No party owns the determination; no one can be held to account.

4
Disregard of Consequences — Intensity 5

No reference in 411 pages to Royal Commission findings, waitlists, service caps, unmet need, misclassification harm, or appeal pathways. Consequences absorbed silently by clients and unpaid carers.

Step 2 — Mechanisms Continued
1
Dehumanisation — Intensity 4

Reduction of identity to category. The Stolen Generations, Forgotten Australians, and Former Child Migrants collapsed to one tick box. Truthful framing: these are people with names, histories, and rights — not data points being binned for budget allocation.

2
Attribution of Blame — Intensity 3

"If a service sub-type is not selected and the client subsequently needs it, the client will need to request this via a SPR." (p. 373) Clients bear the administrative cost of the assessor's or system's omission.

3
Moral Justification — Intensity 4

Document framed throughout as "person-centred", "strengths-based", "empowering" while operating as a rationing instrument. Rationing can be defended ethically — but only with weights and trade-offs published.

4
Advantageous Comparison — Intensity 3

"The IAT builds on the previous NSAF and learnings from the two trials." The IAT inherits the NSAF's structural opacity. The reform narrative pre-empts the critique of continuity in opacity.

Cumulative Threshold & Verdict
Base Rule — Satisfied

Three or more mechanisms at intensity 3 or above: all eight mechanisms register at intensity 3 or higher.

Political Escalation Clause — Fires

Three or more mechanisms at intensity 4 or above under high political context: satisfied by six mechanisms.

The Architectural Triad

Three mechanisms are entrenched at intensity 6 — the triad that protects the algorithm from accountability:

  • Euphemistic Labelling — renders rationing invisible
  • Displacement of Responsibility — insulates the Department
  • Diffusion of Responsibility — prevents any party from being held to account
The Story

The assessor arrives at the small fibro house on the edge of a coastal town. Eighty-three years old, widowed two years ago. Heart failure. A right knee that no longer takes weight. Three pills in the morning, four at night. A daughter in Brisbane. A cat. The garden is overgrown. The shower has no rail.

The assessor works through the IAT. She rates the mobility, runs the DEMMI modified, asks about loneliness, codes the conditions. The older woman says: "I want to stay in my house. I want to be here when my daughter visits at Christmas." The assessor writes that down. She presses the button. Classification 3.

She knows the woman is closer to a 5. The weights are not published. She cannot back this up. She accepts the outcome. Six weeks later the woman falls in the bathroom. There was no rail. The Department does not publish the weights. The Department does not publish what happens to women like this. The system is not in the room.

Priority Interventions
1. Publish the Schema

The Department must publish the IAT's calculation logic — variables, weights, decision rules, validation methodology, and override architecture — under standards equivalent to those expected after Robodebt. The public does not currently have the information needed to judge whether the algorithm is sound.

2. Establish Independent Algorithmic Review

An independent body — at the level of the Inspector-General of Aged Care — should audit the IAT for: training cohort representativeness; disparate impact across diversity fields (CALD, First Nations, LGBTI, care leavers, rural/remote); sensitivity to borderline classifications; and consequences of the override prohibition. The audit must be published.

3. Resolve the Override Contradiction

Either the system should not display the prompt where policy forbids the action, or the policy should permit a documented, evidence-based override with appeal rights for the client. Leaving the contradiction unresolved teaches assessors that the system says one thing and policy says another — and they must accept the system's first determination either way.

Reflective Check & The Undiscussables
Did Safety Constraints Limit This Analysis?

No. The analysis surfaces an institutional architecture; it does not facilitate harm. The strongest claims — the comparison to Robodebt's silhouette, the naming of the override contradiction as moral disengagement at intensity 6 — are warranted by the document itself.

The Undiscussables Surfaced
  • The weights of the algorithm
  • The appeal pathway for ineligibility determinations
  • The disparate impact across diversity fields
  • The political function of opacity in a domain where the Royal Commission has already established the cost of unaccountable systems

Each is a legitimate object of public inquiry. Each is suppressed by the architecture of the user guide.

Closing
Source

My Aged Care — IAT User Guide, Version 2.4, November 2025. Department of Health and Aged Care.

Instrument

Deep Truth Persona Version 5.2 (consolidated), 17 May 2026.

Verdict

Red — cumulative pattern of moral disengagement under high political context. Political Escalation Clause fired.

"Listen to the world."
— instruction from a Spanish grandfather, carried into a method.

Deep Truth | Mindful Progress | Steve Davies | May 2026
Published under Creative Commons