BEE: Belief-Value-Aligned, Explainable, and Extensible Cognitive Framework for Conversational Agents

Sep 16, 2025·
C. Yang
,
M. Gross
Dr. Rafael Wampfler
Dr. Rafael Wampfler
Abstract
We present BEE, a Belief-Value-Aligned, Explainable, and Extensible cognitive framework for conversational agents. BEE decomposes agent cognition into explicit, modular components: a persistent belief management system that maintains the agent’s world model; a value alignment layer that constrains response generation to conform with the agent’s ethical commitments and personality; an explainability mechanism that exposes transparent reasoning traces for auditing and debugging; and an extensible architecture that supports new knowledge domains and character profiles without retraining. Together, these components enable agents that are both technically principled and interpretable. We evaluate BEE on extended multi-session dialogues with embodied agent characters, demonstrating improved consistency and value alignment compared to baseline approaches.
Type
Publication
In Proceedings of the 25th International Conference on Intelligent Virtual Agents (IVA ‘25), Berlin, Germany
publications

Best Paper Honorable Mention — 25th International Conference on Intelligent Virtual Agents (IVA 2025)

Cognitive Architectures Conversational Agents Explainability Value Alignment Belief Management
Dr. Rafael Wampfler
Authors
Dr. Rafael Wampfler
Senior Researcher & Lecturer

I am a Senior Researcher & Lecturer at the Computer Graphics Laboratory of ETH Zurich, and a Research Consultant at Disney Research. I am leading the Digital Character AI projects at CGL. My research interests include conversational digital characters, affective computing, human-computer interaction, and applied machine learning.

My vision is to create intelligent digital humans that can naturally communicate, understand, and support people across domains such as education and mental health. My research focuses on multimodal artificial intelligence for interactive digital humans, developing models that combine large language models, affective computing, and data-driven animation to create embodied conversational agents endowed with autonomous agency, consistent values, and beliefs.

My work bridges machine learning, human–computer interaction, and computer graphics to enable AI systems such as Digital Einstein and interactive patient avatars for psychotherapy training and health education.