Academic AI overreliance scale: development and preliminary psychometric validation in health sciences university students.

The integration of generative artificial intelligence (GAI) tools into higher education is profoundly transforming learning processes (, ), particularly in disciplines related to the health sciences (). Unlike earlier forms of artificial intelligence primarily focused on classification or prediction, generative systems are capable of producing original natural-language content, including academic summaries, conceptual explanations, clinical analyses, and complex argumentative texts. These platforms can synthesize evidence, structure reasoning, and solve demanding tasks in real time, thereby altering the ways students access, organize, and process academic information (, ). This phenomenon has generated growing interest within the field of medical and health sciences education, where debates are beginning to emerge regarding the potential benefits and risks of these technologies for professional training (). In health sciences, the integration of generative AI is especially relevant because academic training depends not only on the acquisition of knowledge, but also on the progressive development of clinical reasoning, professional judgment, and decision-making under conditions of uncertainty (). Current discussions surrounding the use of generative AI in education appear increasingly polarized, resembling the positions described by Umberto Eco in his debate between the “apocalyptic” and the “integrated” perspectives on mass culture (). From an “integrated” perspective, AI is portrayed as a tool capable of democratizing access to knowledge, optimizing personalized learning, and improving academic efficiency (, ). Within this framework, AI functions as an advanced cognitive support system that amplifies students’ intellectual capacities and facilitates adaptive learning processes (). In contrast, from a more “apocalyptic” standpoint, some scholars warn that intensive AI use may erode fundamental cognitive skills, diminish critical thinking, and promote forms of intellectual automation incompatible with deep learning (, ). These concerns are particularly salient in health sciences programs, where clinical reasoning fundamentally relies on the ability to integrate evidence, evaluate ambiguous information, and sustain complex decision-making processes. Within this context, particularly in the fields of psychometrics and digital behavior assessment, attempts have begun to emerge to operationalize forms of intensive or dysfunctional AI use. Concepts such as “AI addiction” (), “problematic AI use” (), and “AI dependency” () have recently appeared in the literature as researchers attempt to adapt frameworks derived from problematic technology use and behavioral addictions to interactions with generative AI systems (). However, traditional models of problematic technology use are generally grounded in mechanisms associated with hedonic gratification, reward-seeking, behavioral compulsion, and variable reinforcement (, ). Platforms such as social media and online video games primarily operate through dynamics designed to maximize emotional engagement and prolonged behavioral involvement (). By contrast, generative AI in academic contexts appears to follow a different functional logic. Students do not use these tools exclusively for entertainment or emotional regulation, but also to delegate complex cognitive tasks directly linked to learning and academic performance ().