This paper introduces a mathematical framework for the formal modeling of human value structures, aimed at mitigating insider threats in information security. A modified version of the Rokeach Value Survey is developed, encoded as a semantic differential, enabling the quantification of subjective value judgments. Principal Component Analysis is then employed to extract the dominant dimensions underlying individual value systems. These dimensions are embedded within a p-adic coordinate system, resulting in an ultrametric state space where the hierarchical organization of values is preserved, and p-adic distances provide a natural measure of proximity between individual value profiles. To model dynamic changes in mental states, we propose a novel adaptation of Hamilton’s principle of least action within this non-Archimedean framework. This approach facilitates the prediction of trajectories through the value space, potentially reflecting shifts in intention or motivation relevant to insider risk. The integration of p-adic analysis and ultrametric geometry offers a new, quantitatively robust perspective on cognitive and behavioral modeling, providing a foundation for future empirical studies and practical tools in cognitive cybersecurity.