What the seminar covered

Engineering conversational digital twins
Digital twins combine AI models, personal data, voice, visual representation, and real-time interaction into a single system. I discussed how these components can be designed and integrated to create responsive digital representations while maintaining clear boundaries around identity and authenticity.

AI avatars as interactive systems
Avatar interfaces introduce challenges that go beyond conventional UI design. Facial expression, voice, timing, and conversational behaviour all influence how users perceive and respond to an AI system, making interaction design and technical implementation closely interconnected.

Designing for privacy, consent, and user control
Systems built around personal identity can process sensitive inputs such as voice, likeness, behavioural signals, and personal narratives. The seminar explored how consent mechanisms, data minimisation, transparency, and user control can be embedded directly into the system architecture.

Building for sensitive real-world contexts
AI avatars deployed in cultural, educational, or emotionally sensitive environments require more than a working prototype. Designing for these contexts means anticipating failure cases, defining system boundaries, and testing how users interpret and interact with AI-generated representations.

From AI prototype to accountable system
A central focus was how experimental AI and avatar technologies can become robust public-facing systems. This includes traceable system behaviour, understandable AI outputs, reliable interaction flows, and technical safeguards that make deployment possible in high-trust environments.