Universities worldwide are grappling with how to integrate generative artificial intelligence tools into teaching and learning while building students' critical capabilities. A new study by Sylvana Sofkova Hashemi at the University of Gothenburg examines how students engage in multimodal prompting when working with these systems, framing the activity as essential epistemic work that advances AI literacy.
The research, published in 2026, draws on empirical data from higher education settings to reveal the prompting strategies learners develop and the reflective discussions that accompany their experiences with generative AI in everyday study practices. The full publication is available at https://www.sciencedirect.com/science/article/pii/S2666920X26000974.
Understanding Multimodal Prompting in Generative AI Contexts
Multimodal prompting refers to the practice of crafting inputs for AI systems that combine text with other modes such as images, audio, video, or structured data. In higher education, students increasingly use these techniques to summarize research articles, generate concept maps, transform writing styles, or organize lecture notes. The process requires more than technical skill; it involves iterative refinement, evaluation of outputs, and integration of AI-generated material into personal knowledge frameworks.
Sofkova Hashemi's work positions this prompting as epistemic work, meaning it contributes directly to knowledge construction rather than serving as a mere shortcut. Students experiment with prompt variations, assess relevance and accuracy, and adjust their approaches based on feedback loops with the AI. This mirrors broader shifts in academic work where human-AI collaboration shapes how knowledge is produced and validated.
Background on AI Integration in University Settings
Generative AI tools have moved rapidly from experimental status to everyday use in classrooms and independent study. Faculty members report students turning to these systems for drafting, research synthesis, and creative tasks across disciplines from humanities to engineering. At the same time, concerns about over-reliance, reduced critical thinking, and uneven access have prompted institutions to develop guidelines and literacy programs.
AI literacy encompasses the abilities to understand AI capabilities and limitations, interact effectively with systems, evaluate outputs critically, and apply the technology ethically in academic and professional contexts. The University of Gothenburg profile of Sylvana Sofkova Hashemi highlights her long-standing focus on digital media and literacy development in changing text and media landscapes, providing continuity with this latest contribution.
Key Insights from the 2026 Study
The investigation centers on students' prompting strategies and their accompanying reflections. Participants described developing systematic approaches: beginning with broad queries, then layering constraints such as audience, format, or disciplinary conventions. Multimodal elements proved particularly useful for tasks requiring synthesis of visual and textual information, such as creating study aids or analyzing complex datasets.
Reflective discussions revealed how students viewed prompting not as isolated commands but as ongoing dialogues that supported deeper engagement with course material. Many noted that successful interactions depended on their existing subject knowledge, which guided prompt refinement and output assessment. The study underscores that AI literacy develops through active, situated practice rather than abstract instruction alone.
Implications for University Educators
Faculty can build on these findings by incorporating explicit prompting activities into coursework. Assignments that ask students to document their prompt iterations, compare AI outputs with traditional sources, and articulate decision-making processes help surface the epistemic dimensions of the work. Professional development for instructors might include workshops on designing such tasks across disciplines.
Departments exploring curriculum updates may find value in aligning AI-related learning outcomes with existing information literacy frameworks. This approach supports students in transferring skills from one course or tool to another while maintaining academic integrity standards.
Benefits for Student Learning and Career Preparation
Students who master multimodal prompting gain advantages in efficiency and creativity during their studies. They also prepare for workplaces where similar skills support research, content creation, and problem-solving. The epistemic framing emphasizes agency: learners remain in control, using AI as a collaborator rather than a replacement for their own thinking.
Graduate students and early-career researchers may particularly benefit, as these practices align with demands for rapid literature reviews, grant writing support, and interdisciplinary collaboration. Institutions that offer structured opportunities to develop these competencies position their graduates competitively.
Challenges and Considerations in Implementation
Despite the promise, several hurdles remain. Uneven prior experience with technology can create disparities among students. Institutional policies on AI use vary, sometimes creating confusion about acceptable practices. Additionally, the rapid evolution of tools requires ongoing adaptation of teaching approaches.
Equity issues surface around access to premium features or reliable internet connections. Educators are encouraged to provide low-tech alternatives and emphasize transferable strategies that work across platforms. Ethical dimensions, including attribution and bias awareness, warrant explicit attention in classroom discussions.
Future Directions for AI Literacy Research and Practice
Sofkova Hashemi's contribution opens avenues for longitudinal studies tracking how prompting skills evolve over degree programs. Comparative work across institutions and countries could illuminate cultural and disciplinary differences. Partnerships between researchers and technology developers may yield interfaces better suited to educational prompting needs.
Broader adoption of epistemic approaches could influence policy at national and international levels, encouraging funding for literacy initiatives that prioritize critical and creative engagement over purely technical training.
Photo by Markus Spiske on Unsplash
Actionable Steps for Academic Institutions
University leaders can begin by auditing current AI-related offerings and identifying gaps in support for multimodal practices. Pilot programs in select departments allow for iterative refinement before wider rollout. Collaboration with libraries and teaching centers often yields effective resource development.
Regular forums for sharing effective practices among faculty foster a community of inquiry. Student input through surveys or focus groups ensures that initiatives address real needs and experiences.