Towards Human-AI Teaming for Skill Development: From Dyadic Interview Practice to Triadic Programming Collaboration

As the job market for CS graduates grows increasingly competitive, effective preparation for technical interviews through mock interviews and programming practice has become critical. This thesis explores how AI can support skill development in these domains, progressing from AI as a sole practice partner to AI as a collaborative teammate. Our first two studies investigate LLM-based conversational AI for interview preparation. Study 1 developed an interview system grounded in reflective learning and dialogic feedback, enabling learners to engage in low-stakes practice with personalized, interactive feedback. Study 2 extended this work to technical interviews, exploring how conversational AI can support the think-aloud practice in technical interviews through simulation, feedback, and example. Together, these studies demonstrate the value of AI as a dyadic practice partner. However, participants reported that while AI practice was useful, peer-based engagement may offer stronger social connection and motivation, suggesting that AI should augment rather than replace human collaboration. This insight motivated our third study, which investigates human–AI teaming in a triadic configuration. We introduce human–human–AI triadic programming, where two humans collaborate with a proactive AI agent. Results from 20 participants show that this triadic collaboration improves collaborative learning and social presence while encouraging more responsible AI use. Together, these studies advance understanding of how AI can support skill development, illustrating a trajectory from dyadic human–AI interaction toward richer forms of human–AI teaming that preserve the pedagogical and social benefits of human collaboration.