Inferential Density: A Framework for Measuring What Information Environments Make Knowable

Inferential Density: A Framework for Measuring What Information Environments Make Knowable This working paper introduces Inferential Density, an information-theoretic framework for examining how the integration of data sources increases the recoverability of attributes that are not explicitly disclosed. The framework defines inferential density through normalized mutual information and proposes an operational estimation approach based on calibrated, cross-validated predictive log loss. A controlled synthetic proof of concept explores different levels of information integration (E0–E4), illustrating how individually limited signals can become substantially more informative when combined. The paper also discusses exploratory observations involving mobility data and cross-device screen-time telemetry. These cases illustrate the distinction between recorded activity, inferential capability, and verified human behavior, highlighting the risks of measurement ambiguity and overinterpretation. The study distinguishes theoretical formulation, synthetic experimental evidence, and exploratory real-world observations. Historical aggregate results and supporting materials are provided where available, while limitations concerning the reproducibility of the original experiments and the availability of raw mobility data are explicitly documented. The proposed framework contributes to discussions of inferential privacy, behavioral opacity, information environments, data integration, and AI governance. Research status: Independent research working paper, version 1.2 (November 2026). Correspondence: [email protected]