Generative Modeling of Individual Behavior at Scale
This work addresses the need for scalable and generative modeling of individual behavior, which is incremental by building on existing stylometry methods but extends them to new domains and larger scales.
The paper tackles the problem of modeling human behavior at the individual level by framing behavioral stylometry as a multi-task learning problem, using parameter-efficient fine-tuning to learn generative style vectors for each person, and demonstrates its approach on chess (47,864 players), Rocket League (2,000 players), and image generation for 10,177 celebrities, enabling style steering towards desired properties.
There has been a growing interest in using AI to model human behavior, particularly in domains where humans interact with this technology. While most existing work models human behavior at an aggregate level, our goal is to model behavior at the individual level. Recent approaches to behavioral stylometry -- or the task of identifying a person from their actions alone -- have shown promise in domains like chess, but these approaches are either not scalable (e.g., fine-tune a separate model for each person) or not generative, in that they cannot generate actions. We address these limitations by framing behavioral stylometry as a multi-task learning problem -- where each task represents a distinct person -- and use parameter-efficient fine-tuning (PEFT) methods to learn an explicit style vector for each person. Style vectors are generative: they selectively activate shared "skill" parameters to generate actions in the style of each person. They also induce a latent space that we can interpret and manipulate algorithmically. In particular, we develop a general technique for style steering that allows us to steer a player's style vector towards a desired property. We apply our approach to two very different games, at unprecedented scales: chess (47,864 players) and Rocket League (2,000 players). We also show generality beyond gaming by applying our method to image generation, where we learn style vectors for 10,177 celebrities and use these vectors to steer their images.