Impact of Artificial Intelligence on Research Practices in Pakistan: A Systematic Analysis of Emerging Trends

Abstract

Generative artificial intelligence (AI) and large language models (LLMs) have ushered in a paradigm-changing era in the field of higher education. In a country like Pakistan, there are substantial tools used in an academic environment to boost the production of knowledge. There is however a crucial empirical gap still to be filled, concerning the structural change of research workflows in force with AI co-agents in the resource-poor context of intense metrics-driven publishing pressures. This study examines the trends, structures and behaviours of integrating AI in Pakistani universities. This study employed a mixed-methods approach using quantitative survey responses (N=400) analysed through the use of PLS-SEM, with some qualitative thematic findings obtained from twenty-five semi-structured faculty interviews. Results show that the relationship between Perceived Usefulness and AI Adoption Intensity has a significant value (β = 0.492, p < 0.001) where Perceived Usefulness has a direct impact on productivity. There is, however, an important issue with every literature mapping and manuscript polishing by AI: AI hovers the prospect of cognitive dependency, exposure to algorithmic hallucinations in literature, and a potential loss of original conceptual syntheses. In practical terms, such ideas call for a paradigm shift in the policy vacuum higher education sector of Pakistan has been facing, to start using the disclosure systems and assessment on quality parameters. This study goes beyond the conventional acceptance models by addressing hidden cognitive costs of human-AI collaboration, in the Global South.