Abstract: Large Language Models (LLMs) and other foundation models attempt to understand sequences of words, matrices of pixels, or timelines of audio spectra largely in isolation from the people generating them. In this talk, I will discuss work that reconceptualizes language modeling as modeling human behavior and minds, updating the foundational task of predicting the next word to also account for the person producing it. Drawing on psychological theories, our human language modeling framework explicitly represents relatively stable individual differences and dynamically changing states. I will present evidence that these models improve performance on traditional natural language processing tasks, as well as for assessing psychological and mental health outcomes. Taken together, this work illustrates a vision in which modern psychology and human-centered AI develop symbiotically, improving AI models while also deepening our understanding of the human condition.