Power-Full by Design: How AI Can Enhance Experimental Research Methods

Artificial Intelligence (AI) is increasingly used in research, yet we know little about how to integrate it effectively and transparently into social science experiments. Using one of my working papers as an illustrative example, I discuss how AI can streamline and advance the methodological aspects of experimental research in three ways. First, I explain how researchers can design simulation-based power analyses that account for parameters and assumptions often overlooked by baseline tools such as G*Power. Second, I talk about how Large Language Models (LLM) can be used to explore potential causal mechanisms between key variables of interest. Finally, I reflect on employing AI-generated video to model dialogical conditions in experiment treatments. All researchers interested in AI or experimental methods are welcome to join the discussion.