Navigational Epistemology: Toward a New Understanding of Knowledge in the Age of Generative AI

Generative artificial intelligence systems have demonstrated capabilities that challenge existing conceptual frameworks used to describe their operation and interaction with users. Current interpretations often emphasize tool use, collaboration, or statistical learning, but may overlook the role of the underlying representational space in shaping system behavior. This paper examines generative AI through the lens of high-dimensional geometric representations. Building on connections to threshold logic and early neural models such as the perceptron, the analysis explores how properties of high-dimensional spaces—such as concentration of measure, near-orthogonality, and manifold structure—affect representation and generalization. These properties are interpreted in epistemic terms, suggesting that meaning in generative systems emerges from positional relations within learned representations. Based on this perspective, the paper introduces the notion of navigational knowledge, describing interaction with generative models as movement within a structured representational space. The paper proposes navigational thinking as an epistemic paradigm in which navigation within representation spaces complements and, in some cases, precedes computation.