Exploring the impact of multimodal generative artificial intelligence on CT dispositions in inquiry-based academic writing: a quasi-experimental analysis in higher education

Introduction Generative artificial intelligence (GenAI) offers new possibilities for scaffolding inquiry-based writing (IBW). However, its impact on critical thinking (CT) dispositions remains empirically underexplored, particularly regarding how the modality of GenAI interaction shapes this impact. Grounded in the Cognitive Theory of Multimedia Learning, this study examines whether multimodal GenAI, which simultaneously processes and generates text, images, and visual representations, cultivates CT dispositions and improves academic writing quality more effectively than text-based GenAI or no GenAI support. Methods A quasi-experimental pretest-posttest design with three conditions (multimodal GenAI, text-based GenAI, and no GenAI control) was implemented with 165 undergraduate students over an eight-week IBW intervention. One-way ANCOVAs were conducted to compare condition effects on posttest outcomes, controlling for pretest scores. Results Analyses revealed significant main effects of condition on both CT dispositions and academic writing quality. The multimodal GenAI group showed the greatest improvement on both outcomes, followed by the text-based GenAI group, with the no GenAI control group showing the smallest gains. Subdimensional analyses indicated that the multimodal advantage was concentrated in openness and analytical thinking within CT dispositions, and in evidence use and analytical depth within writing quality. Discussion These findings extend the Cognitive Theory of Multimedia Learning into GenAI-assisted learning contexts and demonstrate that the modality of GenAI interaction is a consequential design variable for cultivating CT dispositions through IBW. Practical implications for educators and instructional designers are discussed.