Not Just Correct: Exploring Empathetic AI Tutor Design for Game-Based Programming Interfaces

This study explores how learners with different emotional and motivational profiles respond to AI tutor strategies such as scaffolding, humor, and avatar design in a GenAI-based, game programming context. A pre-study and scenario-based survey (N = 30) assessed responses in terms of empathy, clarity, and motivation. Scaffolded feedback was most effective, especially for anxious learners. Affil-iative humor and expressive avatars boosted engagement, while directive and neutral styles were less helpful. A configurable chat-bot prototype was developed and integrated into U-Sharp, a Unity-based learning game. Results highlight the need to balance empathy, transparency, and learner autonomy in emotionally adaptive GenAI tutor design.