Knowledge-Based Design Requirements for Generative Social Robots in Higher Education

GSRs need multiple types of knowledge to act responsibly: The research identifies three critical knowledge categories—self-knowledge (personality traits), user-knowledge (student learning goals and emotional state), and context-knowledge (course materials and teaching strategies)—as prerequisites for generative social robots to function ethically and effectively as tutors, rather than simply following general responsible AI principles.

Personalization is essential for preventing overreliance and misinformation: By understanding individual students' learning progress, motivation types, and backgrounds, tutoring robots can provide tailored guidance that reduces risks of students blindly accepting incorrect information or becoming overly dependent on the system.

Current responsible AI frameworks fall short for educational applications: The paper reveals that existing responsible AI guidelines "primarily define desired behaviors, yet they rarely specify the knowledge prerequisites" needed for generative agents to reliably express ethical behavior in real educational settings.

Students and educators have specific expectations for robot personality and boundaries: Based on interviews with university students and lecturers, the research found that tutoring robots should have customizable roles and conscientious personalities that acknowledge their limitations—suggesting that responsible design requires understanding stakeholder expectations, not just technical safeguards.

Context matters for responsible deployment: Tutoring robots need knowledge of the physical learning environment, specific course content, and institutional educational strategies to avoid giving generic or contextually inappropriate advice that could undermine learning outcomes or violate privacy norms.