PhD Research · UTS Data Science Institute

From Compliance Theatre to Capability: Algorithmic Liability and the Evidentiary Gap in AI Governance

PhD Thesis · Human Centred AI Lab · University of Technology Sydney · 2026–2028

Active Research UTS DSI HCAI Lab AI Governance Director Liability s.180 Corporations Act
When a regulator, a court, or an inquiry asks a board to demonstrate what it actually knew, decided, and controlled about an AI system — can it answer? Not with a framework document. With evidence.
The Research
Where This Started — and Where It Led

The research began by asking which governance artefacts would help organisations genuinely adopt Responsible AI. A practical question — the kind a practitioner asks. But as fieldwork developed, the deeper question became unavoidable.

Most organisations can demonstrate governance intent. Policies approved. Risk committees convened. Frameworks adopted. Very few can demonstrate governance execution — what the AI system actually did, under what parameters, at the moment a consequential decision was made.

That gap is the research. Not the frameworks. The distance between what a board says it governs and what it can reconstruct at the moment of accountability.

The Core Concept
The Evidentiary Gap

The Evidentiary Gap is the distance between process evidence and system evidence. Current AI governance frameworks are almost entirely built on process evidence. The legal and regulatory standard is moving toward system evidence. Most organisations do not know how far they are from it.

System Evidence
What the AI actually did — reconstructable, deterministic, auditable
↔
Process Evidence
What governance intended — policies, frameworks, committee minutes
A policy document is not evidence. A committee meeting is not a control. Under s.180 of the Corporations Act, the question is not whether governance existed. It is whether governance can be proved.
Thesis Structure
Three Layers the Research Maps
01
Legal and Fiduciary Standard
The conduct standard under s.180 of the Corporations Act and the 2026 Privacy Act reforms. What a regulator, court, or inquiry is actually asking for — and why process evidence will not survive the next wave of regulatory scrutiny.
02
Technical Architecture
What deterministic telemetry and causal audit trails look like in practice. The engineering layer that makes governance execution provable — not asserted. Grounded in the Causal Quantification Framework developed with Professor Jianlong Zhou.
03
The Structural Gap
Why current frameworks almost entirely ignore the difference between process evidence and system evidence — and what closing that gap requires at the board, legal, and engineering layer simultaneously.
Applied Research Tool
E-Lens and ABE

The applied research instrument for this thesis is a diagnostics framework developed at UTS DSI that generates continuous trustworthiness metrics for AI systems in deployment.

UTS DSI Diagnostics Framework
E-Lens · ABE
Developed at the Human Centred AI Lab. Generates continuous trustworthiness metrics for AI systems operating in real-world deployment. Not a point-in-time audit. A continuous measurement layer that makes the Evidentiary Gap visible — and measurable — in practice.
Published Research
Output to Date
Research Paper · AICD Editorial Style · 2026 Featured · HCAI Lab
Beyond Probabilistic Governance: Why Continuous Auditing Is the Only Legal Defence for Agentic AI
Sam B · AI Decoded · UTS Data Science Institute
Professor Jianlong Zhou · Human Centred AI Lab · UTS
Introduces the Deterministic Risk Architecture (DRA) — an engineering approach that wraps probabilistic AI in a deterministic enforcement shell grounded in Professor Zhou's Causal Quantification Framework. Addresses the legal exposure created when Agentic AI operates without deterministic enforcement boundaries. For Australian directors, this is no longer an operational question. Under s.180 of the Corporations Act, it is personal.
"Ground-Breaking White Paper on AI Governance" — featured as lead research by the Human Centred AI Lab, UTS Data Science Institute, April 2026
Agentic AI Director Liability Continuous Auditing ETSI CABCA NIST AI 800-4 DRA s.180
Download PDF → hcai-lab.org
In Progress
What Comes Next
2026 AI Liability Benchmark
Independent, non-commercial · Australian regulated entities · OAIC / AUSTRAC / APRA exposure
Active
PhD Thesis
From Compliance Theatre to Capability · Supervised by Distinguished Prof. Fang Chen · Assoc. Prof. Jianlong Zhou · Dr George Tian
2026–2028
Causal Quantification Paper
Co-authored with Professor Jianlong Zhou · DRA evidentiary architecture
In Progress
Principal Researcher
Sam B
PhD Candidate · UTS Data Science Institute
Human Centred AI Lab · HCAI Lab
Principal · AI Decoded
Four-time founder
LinkedIn →
Supervisor
Distinguished Prof. Fang Chen
Executive Director · UTS Data Science Institute
Human Centred AI Lab · UTS
Eureka Prize Winner · 300+ publications
Supervisor
Assoc. Prof. Jianlong Zhou
Data Science Institute · UTS
HCAI Lab Lead · AI Ethics & Fairness
Causal Quantification Framework
Supervisor
Dr George Tian
Senior Lecturer · UTS Faculty of Law
AI Regulation · RegTech · International Law
WIPO Arbitrator · Harvard & Oxford Visiting Scholar
Research Status
InstituteUTS DSI
LabHCAI Lab
StatusActive
Published1 Paper
CommercialNone
AgendaIndependent
Visit HCAI Lab →