This paper argues that artificial intelligence (AI) research can be improved by incorporating a deeper understanding of decision processes and relevant process data. In particular, we emphasize insights into how decisions are formed over time, and introduce a computational framework for decision evidence accumulation that draws on research in psychology, neuroscience, and economics. We present a method for systematically integrating this framework into the training and deployment of AI, with a particular focus on improving human-AI interaction. We also discuss the extent to which current multi-agent AI approaches leverage process data and decision models.