Generative AI Dependency and Perceived Negative Effects on Problem-Solving Among University Students: A Cross-Sectional Study

Abstract

The increasing adoption of generative artificial intelligence (GenAI) in higher education has raised concerns about students' reliance on AI-assisted technologies and their perceived effects on independent problem-solving. This study investigated the association between GenAI dependency and perceived negative effects on problem-solving skills or efficiency among university students. A quantitative, cross-sectional research design was employed using secondary survey data from 227 active GenAI users in the United States and Indonesia. A seven-item dependency composite was constructed from Cognitive Preoccupation and Withdrawal measures, excluding Negative Consequences items to reduce direct measurement overlap with the outcome. Data were analyzed using descriptive statistics, reliability assessment, Spearman's rank-order correlation, ordinal logistic regression, and sensitivity analyses. The dependency composite demonstrated good internal consistency (Cronbach's α = .801). Spearman's correlation identified a significant positive association between dependency and perceived negative problem-solving impact (ρ = .387, p < .001). The primary adjusted ordinal logistic regression showed that higher dependency scores were associated with greater odds of reporting negative perceived problem-solving effects (adjusted OR = 2.946, 95% CI [2.065, 4.204], p < .001). Withdrawal remained significantly associated with the outcome when examined alongside Cognitive Preoccupation. Exploratory analysis found no statistically significant evidence that the dependency association differed between respondents from the United States and Indonesia (p = .596). The findings suggest that stronger GenAI dependency-related experiences are associated with greater perceived negative problem-solving impact. However, the cross-sectional design and self-reported measurements prevent causal conclusions or claims of objectively measured cognitive decline. The study highlights the importance of responsible GenAI integration and further research using longitudinal designs and objective problem-solving assessments.

Keywords: generative artificial intelligence, AI dependency, problem-solving, university students, ordinal logistic regression, higher education