The A.I.-KEA Effect in Language Learning

Generative artificial intelligence can enable learners to produce sophisticated language, explanations, presentations and written texts with unprecedented speed. Improved output, however, does not necessarily indicate equivalent growth in independently available knowledge or skill. This conceptual paper proposes the A.I.-KEA Effect in Learning as a lens for examining a possible miscalibration between the competence learners perceive after successful AI-assisted creation and the competence they can subsequently demonstrate without comparable support. The construct draws on Norton, Mochon and Ariely's IKEA effect, research on cognitive offloading, desirable difficulties, retrieval practice, metacognitive calibration and emerging experimental evidence concerning generative AI in education. It also uses the Socratic critique of externalised knowledge as a historical frame. The paper defines the A.I.-KEA Effect as a metacognitive miscalibration in which participation in successful AI-assisted creation increases perceived competence beyond what is supported by independently demonstrated competence. Two mechanisms are developed further: borrowed competence, in which AI contributes expert-like features to a joint performance that are not yet independently available to the learner, and a friction paradox in which AI can preserve the psychological rewards of successful creation while reducing some of the cognitive effort traditionally required to produce the same quality of output. Four testable propositions are advanced, together with a practical Own the Output framework based on explanation, reproduction, adaptation and independent use.