Elections are a cornerstone of political accountability in democracies, allowing voters to regularly assess representatives and either re-elect them or vote them out of office. However, fair elections alone are insufficient - voters also need access to objective information about policy contexts and incumbents' records.
The rise of generative AI poses both risks and opportunities for accountability. Technologies like chatbots could flood the information ecosystem with manipulated content that distorts voters' perceptions. But they also offer new ways to provide high-quality, accessible political information.
In this project, we investigate generative AI's dual effects on accountability. First, we examine whether surging misinformation and disinformation from AI systems could impair voters' capacity to hold governments to account. We analyse the volume and characteristics of generative content around elections and model its potential impacts on knowledge, trust, and voting behaviour.
Second, we explore using AI responsibly to enhance accountability. Can tools like interactive chatbots give citizens convenient access to factual, impartial information about policies, economic conditions, and incumbent track records? We design and test AI systems to deliver such information and measure their impacts on accountability.
Bridging computer science and political science, this research provides urgently needed insight into AI's emerging democratic influence. It aims to mitigate risks of manipulation while also developing novel AI applications to make accountability more robust. By furthering theoretical and practical understanding, we can help secure democracy in the age of increasingly powerful generative technologies.