Challenges in Designing Multi-Disciplinary Praxis in AI Policy Education

Over the past few months, I have participated in many conferences and workshops on AI policy, regulation, governance, and related areas. The most interesting thing about AI is that when it comes to risks and policies, everybody has something to say. Everybody has something to say about AI risks, the future of AI, how AI is creating more AI, how AI is taking away jobs and so on. As a researcher, I consider this a valuable opportunity to conduct research by reading reports, conducting interviews, attending workshops, and similar activities to understand how different people view AI. Two of my articles, one in Science and Public Policy and the other in the Journal of Asian Public Policy, were the result of such research work. However, as a teacher, I face several challenges that bother me, which led to a call for a paper abstract, and this blog addresses the motivation for this call.

My first session in the second iteration of the course “Challenges of AI Technology Regulation” began by outlining the policy cycle model and explaining the stages of the policy process. I am not naïve enough to think that in real-world situations, public policy unfolds in linear, circular, or other fixed patterns; the policy cycle framework is the best way to understand and address the wickedness inherent in real-world policy issues and challenges. Remarkably, I can see my students, including those who have graduated, providing feedback that this is useful for understanding real-world problems in their professional roles.

While the construct(s) AI governance, AI regulation, AI regulatory challenge, AI risk(s), etc., can be viewed through a diverse lens: from economics, law, political science, engineering and so on, and some of their scholars trying to hijack the entire conversation, public policy as an interdisciplinary field provides the necessary arena for all these subject experts to engage in a friendly discussion and contribute to the body of knowledge. Each of these disciplines provides a non-redundant but insufficient role in shaping the AI policy discussions in the Public Policy curriculum. However, this is easier said than done when designing and teaching AI policies for postgraduate students, given the many challenges that must be systematically explored.

These challenges can be procedural, including a lack of faculty expertise, an overemphasis on entry-level, job-specific skill requirements, and many others. A more systemic issue is the very nature of AI technologies, which span sectoral boundaries, are fast evolving and require sector-specific expertise to assess risks and regulatory challenges. 

Do you really need sectoral expertise? Did my experience working with a construction company help me to formulate ideas and arrive at arguments? The answer is yes. But is it possible for someone to contribute to the body of knowledge on AI regulation in construction without sectoral expertise in construction? The answer is again yes. In Public Policy schools, we don’t teach AI regulation in isolation; there are other courses, and for a course on AI regulation, the theoretical framework in public policy is a prerequisite. We also teach them social science research methods, in which they learn to collect data from experts with in-depth sectoral knowledge rigorously. 

The real question, then, is whether it is possible to develop sufficient expertise to interpret the AI application and its associated risks. Here, the answer is No. AI, as a technology, is evolving rapidly; what you considered AI many years ago would not be called AI today by most of us. The same will be true of the AI we are encountering now. For regulators, this pacing problem presents a series of regulatory challenges, which is the focus of my entire course. 

Designing a multi-disciplinary praxis in AI policy education is challenging. A course rooted in public policy frameworks, in which AI is discussed through a multidisciplinary lens, may solve the problem for a few students. Still, the question remains: why is it difficult to have such inclusive discussions? All of these were the motivation for my call for abstracts for a panel (titled the same as this blog) on the upcoming Indian Public Policy Network Conference, to be held from June 1 to 4.  Instead of merely discussing this in blog posts, it should be empirically studied and presented in conference breakout sessions, thereby contributing to the body of knowledge and ultimately to curriculum development for teaching AI policies.