Externalization by Design: Cognitive Load, Bounded AI, and Process-Level Accountability in Expert Knowledge Systems
TL;DR
EASCI (Experience, Articulation, Structuring, Consolidation, Innovation), a five-stage framework that separates cognitive tasks and restricts AI involvement, is introduced, to mitigate working-memory overload.
Abstract
The externalization of tacit knowledge often fails for two primary reasons. First, practitioners are required to recall, organize, and present expert reasoning concurrently, which can overload working memory. Second, in the absence of clear procedural rules, generative AI may generate content that appears reliable but lacks adequate validation, source attribution, or systematic review. This paper introduces EASCI (Experience, Articulation, Structuring, Consolidation, Innovation), a five-stage framework that separates cognitive tasks and restricts AI involvement. To mitigate working-memory overload, the framework employs stage separation, uses multiple information-capture modalities, and incorporates a counter-hypothesis step. Seven design principles, five constraints, a process-level explainability layer, and eleven feedback paths collectively ensure that AI outputs do not acquire authority without human review, thereby enabling process-level auditability. The framework was developed through a design science research methodology, including a scoping review, practitioner interviews, and participatory workshops. Its structure has been formally verified using Lean 4, and quantitative evaluation is ongoing.
