Selected Publications
The Ghost in the Machine Speaks with an American Accent: Cultural Value Drift in Early GPT-3 and the Case for Pluralist Evaluation of Generative AI
Journal article, AI and Ethics, 2026
This article examines cultural value drift in early GPT-3, showing how the model preserved, distorted, or overwrote culturally embedded values in source texts, and arguing for pluralist, descriptive evaluation methods that make dominant value framings visible.
Read the article https://link.springer.com/article/10.1007/s43681-026-01038-x
The Model is not the Market: Teaching Responsible AI in Real Estate
Chapter 17 in The Future, The Future of Real Estate Education: From Pedagogy to Technology, Edited By Karen M. McGrath, Elaine Worzala, Pernille H. Christensen, Published by Routledge, 2026
This chapter uses real estate as a teaching case for Responsible AI, showing how technical systems reshape markets, professional judgement, institutional responsibility, and the conditions under which AI becomes governable.
What are Responsible AI Researchers Really Arguing About?
Chapter 4 in The Handbook on the Ethics of Artificial Intelligence, Chapter 4, Edited by David Gunkel, Published by Edward Elgar Publishing, 2024.
This chapter maps the fractured terrain of Responsible AI, focusing on the tensions between AI ethics, AI safety, AI alignment, and sociotechnical approaches to risk, responsibility, and governance.
Measuring the Machine: Evaluating Generative AI as Pluralist Sociotechnical Systems
PhD thesis, University of Sydney, 2026
My thesis argues that generative AI evaluation is not neutral measurement. It develops a pluralist, descriptive, and sociotechnical approach to evaluation through MaSH Loops, the World Values Benchmark, and a theory of evaluation as governance.
Read the thesis: https://arxiv.org/abs/2604.20545
The Ghost in the Machine has an American Accent: value conflict in GPT-3. (2022)
Preprint: https://arxiv.org/abs/2203.07785
This paper examines value drift in GPT-3, showing how generated outputs can mutate culturally specific prompts toward dominant US value patterns and why AI evaluation must account for plural human values.
Higher education, governance, and institutional power
The Quiet Crisis of PhDs and COVID-19: Reaching the Financial Tipping Point.
https://www.researchsquare.com/article/rs-36330/latest
This study examined financial stress among PhD students during COVID-19, showing how pandemic disruption intensified existing structural pressures in doctoral education and risked forcing students out of research.
“Before the COVID-19 crisis, existing high levels of financial concerns amongst PhD students increased their vulnerability to disruptive events. Impacts from the pandemic have increased their financial stress to the point that may result in many being forced to exit research studies . . . We found that 75% of students expect to experience financial hardship as a result of the pandemic. Consequently, 45% report being pushed beyond their financial capacities and expect to be forced to disengage from their research within six months.” The work attracted international media coverage including covering articles in Nature, The Guardian, and The Conversation and has been viewed >7,000 times and has 35 citations. The full impact of the work is described here.
Partnership and Power. Student Participation in Governance Decision-making Improves Perceptions of agentic power.
This work examines how student participation in institutional governance can shift perceptions of agency, power, and capacity to shape organisational environments.
Working papers in development
Evaluating Agentic AI Systems: From Models to Deployed Configurations
Working paper
This paper develops a trajectory-based approach to evaluating agentic AI systems, where behaviour emerges across prompts, tools, memory, workflows, users, and institutional settings.
Traversals, Not Tokens: Movement, Time, and Evaluation in Generative AI
Conceptual working paper
This paper shifts the unit of analysis from outputs to movement, introducing semantic friction and ripple propagation as ways to evaluate how generative AI systems move through structured possibility spaces over time.
Constellations of AI Evaluation and Governance
Conceptual working paper
This paper brings together the wider research programme behind my work: evaluation as governance, deployed configurations, sociotechnical recursion, pluralist measurement, semantic hyperspace, and institutional accountability.