Research in adaptive hypermedia educational systems has increased with the growth of the Internet. Currently, all adaptive hypermedia educational systems collect information about cognitive style through completion of a questionnaire based on a psychometric test. This direct measure may be intrusive and annoying to a student and makes an adaptive system aligned to cognitive style unavailable for students that have not completed the questionnaire. It is posited that non-intrusive methods for determining the cognitive style of hypermedia system users are needed to maximize the usability, functionality, and goals of adaptive hypermedia systems. This paper offers a new approach for the autonomous computer-based assessment of preferred cognitive style that can support studies in user modeling and human-computer interface domains. It further posits a conceptual model that attempts to determine the preferred cognitive style of an online educational hypermedia user through click-stream analysis of their web-based hypermedia choices and browsing patterns.