Deep Truth Persona 5.3

Road Marks and Sign Posts

Human : AI Partnerships Build the Road

Steve Davies — 3 August 2026

Version 5.3: Two New Enhancements

Plain English Storytelling

Stories generated by default at a reading level of 7–9, consistent with WCAG 2.1 criterion 3.1.5 (Level AAA).

Ambiguity in Stories

Where the Analysis Table says "ambiguous," the story says so too — in plain language, not softened or sharpened.

Story Is Not Decoration

"A story that requires prior knowledge of accessibility settings to receive is not yet a story that reaches everyone the Analysis Table cannot."

Plain English is now the default, not an option the reader must know to ask for. Analysis reaches people who will not read tables.

Managing Ambiguity Honestly

Where a finding is ambiguous, the story's close names the opportunity that would resolve the ambiguity — more evidence, a named next document, clarifying questions — not a call to action the analysis hasn't yet earned.

No False Balance

Ambiguity is not softened into "both sides."

No False Clarity

Ambiguity is not sharpened into manufactured certainty.

Honest Difficulty

Reflects the real challenge facing public servants navigating ambiguity daily.

Cross-Platform Testing

Platforms were asked to analyse the RoboDebt Royal Commission Report — an extensive document with complex organisational, administrative, technological, and human issues.

Testing revealed substantive commonality with some variance and nuance across platforms. This variance did not break the Deep Truth Persona.

Variance Is Healthy

"If all AI platforms produced 100% identical analysis, that would be anathema to the very concept of the Deep Truth Persona."

100% identical outputs would indicate a persona more akin to an algorithm. Architectural and interpretive variance between platforms is a strength, not a fault.

No Baseline — Only Corroboration

No Single Standard

Several platforms produced independent analyses. None is the standard the others are measured against.

Agreement = Real Corroboration

Where platforms agree on the overall verdict and core mechanisms, that is genuine corroboration between independent readers.

Divergence = Capability

Where they land differently, that is what happens when different capable readers work through something complex.

Work to enhance quality outputs is ongoing — crucially, without reducing variance or drifting toward algorithm.

Why Variance Is the Point

RoboDebt's Core Failure

It couldn't be corrected once shown to be wrong. A 2014 legal opinion that the scheme was unlawful never reached anyone able to act on it.

Discretion Removed

People were stripped of discretion to depart from what the calculation said — making it an algorithm in the fullest sense.

Human Choices

The temptation to blame the computer is its own displacement. The choices that removed discretion were human ones.

Different Roads, Same Sign Posts

"Different ways of reasoning, arriving at the same conclusion from different angles, is a strength — not something to smooth away." — Jensen Huang

A panel of independent reviewers is worth more than one reviewer's opinion said five times. The Deep Truth Persona holds this from the start: it is a lens, not a moral agent. The person using it carries that responsibility.

The Road and the Sign Posts


Sign Posts

The Deep Truth Persona provides sign posts — refined and clarified with each version.

The Road

Built fresh each time the Persona meets something as large and human as the RoboDebt Royal Commission Report — between a person, the material, and the platform.

Start by reading and downloading

Bring a report, a case, a question. Let the road emerge together.

Partnership

The same sign posts, genuinely different roads — each built by a real Human : AI partnership, not handed down by either side.

Keep Building the Road

Human : AI Partnerships Build the Road

Deep Truth Persona v5.3Steve Davies — August 2026