Seeing the Forest Through the Traits: A Four-Dimensional Framework for Modeling and Measuring Latent User Traits

Understanding latent user traits is critical for designing adaptive systems that personalize effectively and responsibly. However, most existing methods rely on lengthy psychometric instruments or focus narrowly on observable behaviors, limiting their scalability and psychological depth. In this paper, we introduce a novel framework for modeling user traits using four interpretable, higher-order dimensions—Perceptual, Relational, Exploratory, and Self-Agency. We derived these dimensions from an empirical analysis of 16 latent traits across personality, cognitive ability, and perceptual skill. We demonstrate how this dimensional structure, refined via dimensionality reduction and clustering, enables interpretable user modeling without sacrificing coverage. To support efficient user assessment, we develop adaptive diagnostic questionnaires that follow path-based decision trees, reducing the number of items while maintaining strong internal consistency and classification performance. A path-level inter-survey analysis on the Self-Agency dimension shows high convergence with the original Rotter Locus of Control scale (r = 0.849, p < .001). Finally, in an initial application study, we show how this framework improves the prediction of task effort in data visualization contexts, outperforming a user-agnostic model in both cross-user and cross-visualization generalization settings. Together, our contributions offer a modular, scalable approach to capturing latent user traits for adaptive interaction design.