Dr Naz Güler
Your risk committee may be focused on AI performance. Your governance committee may be focused on AI oversight. But neither view, on its own, tells you whether the transformation underneath is actually under control.
This work is grounded in thirty years of executive and transformational delivery across financial services, central banking and market infrastructure. It is also shaped by doctoral research at UNSW into how AI systems behave when they are placed in human decision-making roles.
My research at UNSW Business School looked at how AI systems trained through reinforcement learning from human feedback behave in decision-making roles. These systems defer to human judgment over algorithmic recommendations. That's the reverse of algorithm aversion, the well-documented tendency for people to reject good advice once they know an algorithm produced it. The deference also depends on the role the system is given: the same system behaves differently depending on what it's asked to do. For boards, the takeaway is practical. Oversight needs to look at how AI systems actually behave in context, not just whether they meet performance targets.
- GAICD, Governance Institute / AICD
- PhD, UNSW Business School — AI governance and human decision-making (thesis, DOI 10.26190/unsworks/31006)
- MBA, Macquarie Graduate School of Management
- BSc (Honours), Psychology — UNSW
- Güler, N., doctoral thesis, UNSW Business School. doi.org/10.26190/unsworks/31006.
- Australian Prudential Regulation Authority, Letter to Industry on Artificial Intelligence (AI), 30 April 2026.