Crisis-induced hybrid learning, cognitive offloading, and generative AI reliance among Pakistani CS undergraduates

Purpose In spring 2026, geopolitical tensions prompted the Pakistani government to mandate full online instruction (10 March–3 April 2026), followed by a hybrid schedule for the rest of the semester. This shift is treated here as an externally imposed crisis context for AI adoption, not as a natural experiment. No pre-crisis baseline or control group was available. The study characterises generative AI (GenAI) adoption patterns and the psychological antecedents of AI dependency among undergraduate computer-science (CS) students during this window. Three hypotheses, grounded in the reviewed literature, structured the analysis. Design/methodology/approach A cross-sectional survey was administered across Pakistani higher education institutions (HEIs) in April–May 2026. Of 360 responses collected, two incomplete records were removed, and a thirteen-criterion data-quality screen was ap-plied, yielding N = 299. Fourteen constructs were operationalised from UTAUT, Cognitive Load Theory (CLT), Self-Determination Theory (SDT), and the AI Anxiety Scale (AIAS). Block-entry OLS regression and bootstrapped mediation (5,000 resamples) tested the hypotheses; ANOVA and Spearman correlations supported descriptive analysis of adoption patterns. Findings Adoption was near-universal (>99%), with 45.2% of respondents reporting ≥ 31% of submitted work directly AI-generated. A three-block OLS regression explained 54.6% of variance in AI dependency. Cognitive offloading was the strongest predictor (β = 0.41, p < 0.001), followed by procrastination (β = 0.23, p < 0.001), extrinsic motivation (β = 0.17, p = 0.002), and intrinsic motivation as a protective factor (β = −0.13, p = 0.013). Bootstrapped mediation confirmed procrastination partially mediates the extrinsic motivation–dependency path (ab = 0.201, 95% BC-CI [0.122, 0.289]). Originality/value To our knowledge, this is among the first studies to survey GenAI dependency during an active government-mandated crisis disruption in South Asian higher education. Contributions include a multi-theory construct battery adapted for a crisis context, a thirteen-criterion response-quality protocol, and evidence that habituated cognitive offloading and extrinsic motivation are the primary drivers of AI dependency in this context.