<![CDATA[Purpose: This study examines the epistemological challenge posed by algorithmic rationality to critical thinking in artificial intelligence-mediated Technical and Vocational Education and Training (TVET). As adaptive learning systems, automated competency assessment, learning analytics, and data-driven qualification frameworks increasingly shape vocational education, the study investigates how algorithmic rationality affects the production and validation of educational knowledge through Habermasian epistemology. Methods: The study employs a qualitative conceptual-philosophical inquiry based on library research, conceptual clarification, comparative theoretical analysis, and argumentative reconstruction. Literature on artificial intelligence in education, TVET, algorithmic rationality, critical thinking, AI ethics, and Jürgen Habermas’s concepts of communicative rationality, validity claims, intersubjective justification, and lifeworld colonization was purposively analyzed. Findings: Algorithmic rationality prioritizes prediction, classification, correlation, optimization, and efficiency in educational knowledge production. In AI-mediated TVET, these processes can shape judgments about competence, certification, readiness, and employability. Yet technical accuracy and efficiency do not by themselves establish epistemic validity. AI-generated knowledge becomes educationally valid only when its factual, normative, contextual, and institutional dimensions are open to communicative testing and intersubjective justification. Therefore, AI assessments and recommendations should remain provisional claims, not final epistemic authorities. Research implications: TVET institutions should embed AI-supported decisions within dialogical review mechanisms that preserve meaningful participation by learners, educators, assessors, institutions, and relevant professional stakeholders. Originality: This study develops a Habermasian framework that reconstructs critical thinking as an epistemic-communicative capacity for evaluating and contesting AI-mediated knowledge in TVET.]]>
