Vibe Coding in Higher Education: A Framework for Critical AI Literacy

Abstract

The accelerated adoption of Generative Artificial Intelligence (GenAI) in programming and academic production activities in higher education has been accompanied by a recurring pattern of uncritical use: students prioritize the tool and the rapid attainment of a “functional” artifact without producing evidence of understanding, validation, and authorial responsibility. This article reports a bibliographic survey conducted in 2025 on vibe coding and GenAI in higher education, focusing on impacts on cognitive load and pedagogical consequences (learning, assessment, integrity, and governance). The synthesis prioritizes 2025 findings describing (i) vicious error–correction cycles without learning; (ii) superficial inspection of outputs and the risk of competence erosion; and (iii) gaps in institutional policies and faculty/student training. In light of Cognitive Load Theory, it is argued that vibe coding may reduce Extraneous Load (effort irrelevant to learning, such as syntax/boilerplate friction) and free resources for Germane Load (productive effort devoted to schema construction), provided that instructional design imposes tasks requiring explanation, testing, and justification. As its primary contribution, the paper proposes an original pedagogical framework expressed as a flowchart, derived from the survey synthesis and modeled on Sweller (1988; 2010), taking as a starting point the “Talking To, Through, and About AI” schema (Woo; Guo; Yu, 2025). The framework renders GenAI use visible and assessable through process-based evidence (intent, prompts, versions, tests, metacognition, and authorship), repositioning vibe coding as a pathway to Critical AI Literacy. Implications for curriculum design, assessment, and integrity policies are discussed, along with limitations and an agenda for empirical validation.