ARTIFICIAL INTELLIGENCE DEPENDENCY AMONG UNDERGRADUATE STUDENTS IN UNIVERSITY OF BENIN, BENIN CITY, NIGERIA

Background: The rapid integration of artificial intelligence (AI) tools into higher education has transformed how students access information, complete academic tasks, and engage with learning. While AI offers significant benefits in efficiency and academic support, growing concerns exist regarding excessive student reliance on these tools, with potential implications for critical thinking, cognitive autonomy, and academic integrity. Despite near-universal AI adoption among students globally, empirical data on the nature and extent of AI dependency among Nigerian undergraduates remain limited. Materials and Methods: A descriptive cross-sectional study was conducted among 588 undergraduate students selected from the Ugbowo campus of the University of Benin using a multi-stage sampling technique. Data were collected using a structured self-administered questionnaire that assessed AI tool awareness and usage, patterns of AI use, and level of AI dependency using the validated 22-item Artificial Intelligence Dependence Scale (AIdep-22), which measures four domains: functional dependence, cognitive dependence, emotional dependence, and loss of control. Factors associated with AI dependency were also explored. Data were analysed using IBM SPSS version 29.0; frequencies, proportions, and means were computed for descriptive statistics, while chi-square tests assessed associations between categorical variables at a significance level of p < 0.05. Results: The mean age of respondents was 22.73 ± 4.65 years, with 52.9% being male. Internet access (99.1%) and smartphone ownership (98.6%) were nearly universal. AI tool awareness was reported by 99.3% of respondents, and 98.1% were active users, with ChatGPT being the most recognised (99.0%), most commonly used (83.2%), and most frequently used (83.2%) tool. The primary academic applications were research assistance (72.3%), summarization of materials (69.7%), and writing support (66.7%), with over half reporting daily use. Regarding usage behaviours, 77.9% reported verifying AI outputs relatively more frequently while 58.8% modified AI-generated content before use more frequently, and 15.0% admitted to unethical use including examination malpractice. In terms of dependency, 49.8% of students exhibited low AI dependency, 40.1% moderate dependency, and 10.0% high dependency. Domain-level mean scores were highest for functional dependence (2.86), followed by emotional dependence (2.55), cognitive dependence (2.50), and loss of control (2.27), with an overall AIdep-22 mean of 2.55, corresponding to moderate dependency. Verification behaviour (p = 0.004) and modification of AI outputs (p = 0.05) were significantly associated with dependency level. The key factors associated with AI dependency were heavy academic workload, high performance expectations, ease of AI access, and fear of making errors. Conclusion: AI tools, particularly ChatGPT, have become deeply embedded in undergraduate academic life at the University of Benin, with near-universal awareness and adoption. Most students exhibited moderate AI dependency, with functional dependence being the most prominent domain. Verification behaviour and output modification were significantly associated with dependency level, underscoring the importance of critical engagement with AI-generated content. These findings highlight the urgent need for institutional guidelines, AI literacy programmes, and pedagogical strategies that promote responsible AI use while preserving students' intellectual independence and academic integrity.