From Ai Explanations to Independent Knowledge
Year: 2026
Type: preprint
Abstract: Generative artificial intelligence has dramatically increased access to explanations across scientific, educational, and professional domains. However, receiving an accurate and coherent explanation does not necessarily result in durable, independently usable knowledge. Individuals may recognize and reproduce AI-generated explanations while remaining unable to reconstruct their underlying reasoning, evaluate their validity, or apply their principles to unfamiliar problems. This article examines the distinction between explanation accessibility and independent understanding and proposes the Independent Knowledge Conversion Framework (IKC), a conceptual model describing how externally generated explanations may become internalized, critically evaluated, and transferable human knowledge. The framework integrates established findings from self-explanation research, generative learning, retrieval practice, metacognition, and emerging investigations of AI-supported cognitive offloading. Five interrelated processes are proposed: conceptual decomposition, independent reconstruction, epistemic verification, contextual transfer, and metacognitive recalibration. The article differentiates immediate AI-assisted task performance from autonomous knowledge acquisition and identifies measurable indicators of successful knowledge conversion. It further develops testable propositions and outlines an experimental research agenda for evaluating the framework across educational and professional contexts. The central argument is that the educational value of generative AI should be assessed not only by the quality of explanations it produces but also by the degree to which individuals develop knowledge they can reconstruct, justify, revise, and apply without continued dependence on the originating system.
Author: Shaban, Roman
Language: English
Subfield: Experimental and Cognitive Psychology
Field: Psychology
Domain: Social Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W7227128665
