Designing for Abductive Meaning-Making: An Explainable Interface for GenAI-Mediated Ill-Defined Problem Solving

Sammanfattning

The growing integration of GenAI in education creates a tension between efficient output production and learners’ ability to understand the reasoning behind AI-generated responses, particularly in ill-defined problem contexts. This study reframes explainability not as an intrinsic property of AI systems, but as an interactional process emerging among learners, AI outputs, and interface structures. Drawing on Peircean semiotics, it conceptualizes abductive reasoning as a key mechanism through which learners interpret and reconstruct the meaning of GenAI outputs under conditions of epistemic opacity. Adopting a Research through Design approach, the study develops an Explainable Interface prototype, the Abduction Journal, designed to externalize abductive reasoning through proposition-based interaction and relational structuring. Three participatory design workshops involving 11 participants informed three design principles: acknowledging interpretive instability, expanding the space of interrogation, and grounding AI outputs in problem context. These principles were translated into an interactive canvas and evaluated through user testing with five participants. The findings suggest that predefined propositions can lower the threshold for entering reasoning activity, while relational judgments between propositions remain difficult for users to construct. A tentative pattern was observed between the coherence of participants’ reasoning structures and the quality of their subsequent problem reframing. The study contributes an interface-oriented account of explainability grounded in abductive meaning-making and proposes abductive reasoning as an additional competency dimension for Computational Thinking in GenAI-mediated ill-defined problem contexts.