Enhancing Team-Based Learning through AI-Mediated Communication in Programming Education

As software development accelerates toward “vibe coding”, where natural‑language interaction with AI coding agents increasingly replaces manual coding, communication competence has become as critical as technical mastery. Yet most programming pedagogy still positions AI as a productivity booster rather than a communication mediator that can scaffold collaborative reasoning, role clarity, and equitable participation. This project proposes, implements, and evaluates an AI‑mediated, team‑based learning (TBL) framework that reconceptualizes human–AI collaboration as human–agent–human triadic teaming. Our goal is to help beginning programmers develop three outcomes essential for AI‑first workplaces: (1) effective collaboration across technical and non‑technical roles; (2) critical evaluation of AI‑generated code; and (3) precise articulation of technical concepts for mixed audiences. We embed this intervention in CB2240 Introduction to Business Programming in Python and design it for scalability to advanced IS courses after validation.Pedagogically, the framework integrates Human‑Agent Teaming (HAT) and Communication Accommodation Theory (CAT) to align coordination and discourse in novice teams. We operationalize these theories through three mutually reinforcing mechanics delivered on an AI coding platform: (i) Owner–Stakeholder role rotation, where each student alternates between “Technical Owner” (steering agent execution) and “Business Stakeholder” (framing requirements), promoting equitable leadership and accountability; (ii) a private “My Agent” sandbox that lets students iteratively test prompts, interrogate model outputs, and rehearse explanations before team discussion, shifting habits from heuristic acceptance to analytic evaluation; and (iii) Collaboration History that transparently logs human–AI exchanges, enabling process‑oriented feedback, instructional guidance, and research analysis of participation and discourse patterns. The intervention is delivered over four weeks in a low‑stakes assignment setting with minimal disruption to normal grading.Methodologically, we adopt a mixed‑methods framework aligned with the project’s three research questions (RQ1–RQ3):RQ1: How does structured role rotation support equitable participation among students with different programming backgrounds?RQ2: How do students develop from heuristic acceptance of AI outputs to analytical evaluation through AI‑mediated interactions?RQ3: How does AI‑mediated collaboration enhance students’ ability to articulate technical concepts and programming vocabulary?Outcome assessments include pre/post measures of programming self‑efficacy and AI literacy test, plus AI‑free performance tasks. The study begins with pilots and is implemented across two semesters in CB2240 Introduction to Business Programming in Python (about 150 students). Participation is voluntary, with no disadvantage to non‑participants.We expect the framework to (a) broaden participation by giving every student structured leadership opportunities; (b) improve judgment about when and how to rely on AI through sandboxed critique and team negotiation; and (c) strengthen technical communication, enabling students to translate between business needs and executable specifications. Beyond immediate course benefits, the project produces reusable artifacts (student guides, reflective prompts, analytics rubrics) and an evidence‑based blueprint for scaling to other IS courses (e.g., IS4435 Analytics Programming with Python, IS4834 Business Intelligence and Analytics, and IS4861 Machine Learning for Business). Findings will inform departmental practice, university workshops, and scholarly dissemination, contributing a replicable model for AI‑mediated teamwork that centers human accountability and communicative competence—capabilities that will define professional success in AI‑enabled organizations.