This workshop introduces a theoretical model for understanding multi-agent communication (Cohen and Levesque) in educational contexts where humans and AI co-construct meaning through dialogue. Building on Vygotskian sociocultural theory, Bandura’s agentic perspectives on self, and Searle’s speech act theory, the proposed model conceptualizes illocutionary work as a distributed phenomenon—where interpretive effort, responsibility, and meaning-making are shared among learners, teachers, and AI interlocutors. Rather than viewing student talk as discrete input/output units, the model frames classroom discourse as a throughput system—capturing the rate and flow of interpretive work over time—shaped by role-based preparation (e.g., literature circles, mock trials, debate), where human and AI scaffolding work in tandem to support human skills in mediated interpretation tasks. Illocutionary work, in this sense, represents the “cognitive labor” performed at the level of interaction—not only by individual agents, but as emergent from group-level coordination—and can be broken down into pre-locutionary, locutionary, illocutionary, and perlocutionary work. Examples of how these concepts were applied to the configuration of tasks, roles, and custom GPTs will illustrate their utility. The workshop also proposes preliminary constructs—such as turnshare, floorshare, and human-AI bandwidth ratios to operationalize group throughput in future research and system design. Applying the theory of illocutionary work to specific pieces (e.g., custom GPTs) as well as broader scaffolding systems offers a more principled foundation for educational AI integration. Attendees will leave with a conceptual map of the model, a sample tagging framework, and custom GPT configuration examples for applying these ideas to their own contexts.

Authors:
Paul Sevigny, Ritsumeikan Asia Pacific University, Japan

About the Presenter(s)
Dr. Paul Sevigny is currently a professor at Ritsumeikan Asia Pacific University, in Beppu, Japan. He is currently researching the impact of custom GPTs on CEFR B1+ learners’ transition from discussion to Debate.

Connect on Linkedin
https://www.linkedin.com/in/paul-sevigny-0b556113/

Connect on ResearchGate
https://www.researchgate.net/profile/Paul-Sevigny?ev=hdr_xprf

See this presentation on the full schedule – Monday Schedule

Bookmark