Contextualizing Trust: Designing Transparent Generative Artificial Intelligence Systems for Collaborative Learning

Abstract

Purpose. Students’ trust in generative artificial intelligence (GenAI) learning applications is critical to determine the adoption and use of these systems. The purpose of this study is to explore students’ trust in GenAI and how to design for trust in collaborative learning tasks. We prototyped different interface designs to promote trust in GenAI, with a focus on transparency to make the system’s design, logic, intended use, and limitations accessible to students. We drew from human-computer interaction research to explore different designs: documentation, evaluation, explainability, use contexts, and disclaimers. Design/methodology/approach. We conducted focus groups with 22 graduate students in two GenAI-assisted collaborative learning tasks: ideating design ideas in groups and generating peer feedback. We used thematic analysis to explore themes around (1) students’ experience with and trust in GenAI applications in education, and (2) feedback on different designs to establish transparency. Findings. Findings revealed the importance of contextual and social transparency. Participants highlighted the need to communicate the purposes of using AI, encourage self-driven exploration of the technology, and embed instructors’ demonstration of GenAI use for specific learning tasks. Participants with less experience with the technology frequently emphasized the role of disclaimers, to encourage students to be critical of GenAI output. Originality. Prior research on trust and transparency has focused on technical aspects, such as providing documentation and explaining how the technology works. Findings indicate that the design of GenAI learning systems should also account for contexts, instructional dynamics, and students’ technology competence.