Lost in the Loop: The Limits of a ‘human in the loop’ for AI Governance
In our second virtual lightning talk for International Access to Information Day (IAID) 2022, Jake Goldenfein, Senior Lecturer at Melbourne Law School, will present Lost in the Loop: The Limits of a ‘human in the loop’ for AI Governance.
As data-driven decision systems proliferate, the ‘human-in-the-loop’ has become a central tool in technology regulation. A growing body of literature, however, has begun to challenge the utility of human oversight for automated decisions. In the context of machine learning assisted (i.e. hybrid) decisions, there is almost no empirical evidence suggesting that a human in the loop improves decision quality. So why do policy makers insist on it as a governance tool?
Our intuitions around the role a human performs in an automated or hybrid decision context now need re-evaluation. The ‘human’ in the loop brings with it a normative and political framework associated with the rule of law that increasingly collides with the realities of automated decision-making. As formerly human tasks and decisions are automated, with cognition outsourced to machines and humans re-tasked as monitors and error-correctors, our understanding of the human as an accountable political actor becomes increasingly awkward. This talk explores how, without closer attention to the realities of computationally assisted decision-making, the ‘human in the loop’ risks becoming a mere symbol or talisman of AI governance that helps us avoid the hard questions of accountability, rather than something functional or meaningful.
Date: Thursday 29 September 2022
Time: 2pm - 2.45pm
Where: Online - Microsoft Teams
Audience: This event is open to everyone