A multimodal social semiotic rethinking of agency in AI-assisted designs for language learning and assessment: Examples from Duolingo

Abstract

Abstract This article addresses issues of design and agency, focusing on Artificial Intelligence (AI)-generated tools for speaking practice and assessment in the field of English language learning. It uses examples from the educational technology company Duolingo as agentive designs for enabling and assessing learning. Drawing on multimodal social semiotics (MMSS) as a theoretical and methodological lens, the article analyses selected screenshots from the Duolingo examples as multimodal texts-in-action, showcasing how these AI-assisted learning designs make visible the agency of the designer and the role of AI, and how the design facilitates the user’s agency. In these designs, AI is described as enacting a form of devolved agency, given that it interacts with the user as a proxy for the human designers who shape the affordances of the AI-assisted learning design. Among the implications of the study is an identified need to make more visible the limitations and potential of MMSS theory to account for the role of AI in contexts of learning and communication.