Artificial Intelligence (AI) is increasingly influencing the conditions of the society we live in (e.g., how much people seek and help one another through hard times, how many good jobs exist). My group's research seeks to improve such influences from an AI application designer's perspective: How can we create AI things that improve not only the conditions of individual users and particular communities, but the properties of society as a whole?

In this talk, I will start by sharing our projects on improving generative AI chatbots' influence on public mental health as a case study for this question. Based on these projects, our studies of industry practice, and lessons from history, I propose a new way of thinking about "designing AI for societal good," namely "designing AI as planting seeds" (of better business models, policy, and politics). I close by discussing our early work exploring this design practice and research agenda.