Cognitive dependency on generative AI

Generative artificial intelligence (AI) tools such as ChatGPT, Google Gemini and Microsoft Copilot are now part of everyday academic work for undergraduate students, offering rapid explanations, summaries and drafts. This convenience raises a concern: frequent and uncritical use may encourage cognitive dependency, in which students routinely hand over reasoning, memory and decision-making to a system and lose confidence in working without it. Existing evidence is promising but fragmented. Some studies report learning gains with AI support, while others report reduced effort, weaker critical thinking and poorer unaided performance. This paper presents a proposed mixed-methods study, not a completed one. Its aim is to examine how generative AI usage frequency, cognitive offloading, AI literacy and metacognitive monitoring relate to critical thinking, independent problem-solving and learner autonomy among undergraduate students. The proposed methodology combines a five-point Likert questionnaire for about 150 students, a paired unaided versus AI-assisted problem-solving task, and semi-structured interviews, analysed through reliability testing, correlation, regression, moderation and thematic analysis. The paper contributes a literature synthesis based on verified sources, a conceptual framework with seven testable hypotheses, a measurable evaluation framework with a Cognitive Dependency Index and an Independence Gap, blank result templates and practical recommendations. It argues that AI is not harmful in itself; the educational goal is cognitive independence, so that students use AI to extend their own thinking rather than replace it.