AI poses major challenges to and causes uncertainty for leaders of governments, since they are progressively confronted with the consequences of algorithmic inequality, reinforced totalitarianism, oligopolistic market structures, labor displacement, or political unrest on national or global scale. The need for policy guidance is therefore substantial.
In this project we try to identify commonly used narratives in governmental AI policies. Exploring narratives is an integral part of public policy research. The analysis of policy narratives can provide insight into a current political issue, shed light on the policy context, clarify the main actors and key arguments for and against the issue, and present scenarios of possible future developments. The project thus aims at discovering the contemporary topics and applied rhetoric of governments to either belittle the risks or exalt the opportunities of AI.
It is obvious that there are human capacity limitations for processing all this information. Instead of applying systematic search strategies or coding procedures for organizing and confining the source materials to a volume that is manageable for humans, we used machine learning for analyzing the entirety of information and for exploring the hidden thematic structures across the distinct policy documents. Though, to make the machine learning output interpretable and meaningful for humans, we combine computational and policy-related methods and distill the narratives which are common in extant governmental AI policies.
Project by: Tobias Mettler, Ali Asker Gündüz
