Faculty of Health, Medicine and Behavioural Sciences
Availability:
Available for supervision
Media expert
I study the risks from advanced AI and the policies that could reduce them. As an Associate Professor of Psychology at UQ and an Affiliate Researcher at MIT FutureTech, I run large, transparent studies that help governments and the public understand where AI is heading. My team built the Survey of AI Risk (SARA), the largest study of how Australians perceive AI, and co-authored the MIT AI Risk Repository, a public catalogue of AI hazards cited in the International AI Safety Report (2024) and Australia's proposal for mandatory AI guardrails.
I make this analysis rigorous, but translate it in ways people can understand. For over a decade I have built systematic reviews, meta-analyses, and structured expert elicitations, including a Delphi study of 272 experts mapping AI risk across 24 subdomains. My current work grades frontier labs' safety frameworks against emerging law, tracks dangerous-capability progress, and benchmarks AI risk against the safety standards other industries already accept. I also chair Effective Altruism Australia, which directs over AU$7.5 million a year to cost-effective global programs.
Research interests
Reducing catastrophic risks from AI. Four priority risks drive the work: sudden loss of control, gradual disempowerment as decisions are delegated to AI, concentration of power, and misuse by rogue actors. I focus on which mitigations experts agree on and what would tell us a risk is rising.
Mapping and measuring AI risk. Through SARA and the AI Risk Repository, I identify hazards, track public attitudes, and build evidence policymakers can use.
Evidence synthesis and expert elicitation. Systematic reviews, meta-analysis, and Delphi methods: how to produce trustworthy evidence at the speed frontier AI demands.
