About
Human communication, psychology and software engineering in one agent lab
Agent Interaction Lab brings together software engineering, cognitive psychology, speech interaction research, non-verbal communication and creative prototyping. The site explores how AI agents can become more than answer machines: they can become communication partners, solution partners and design partners.
What this site explores
Agents as communication partners, design partners and solution partners
The three demos show different directions. Eva is a communication coach that can hold a short dialogue and reflect on interaction patterns. The Multi-Agent Solution Designer combines human psychology and software architecture perspectives into a practical solution plan. The 3D Designer turns prompts into early printable design concepts and a next-step design package.
The work is shaped by Pashiera Barkhuysen's background in audiovisual prosody in interaction, human-agent communication and practical software development. That combination gives the lab its focus: build things that are technically useful, but also sensitive to how people actually talk, hesitate, repair misunderstandings, respond emotionally and collaborate.
Adaptive agents
Each agent tunes in to how you talk
Each agent tunes in differently, adapting to your language, tone, conversation signals and the interaction mode you choose, so every reply fits the moment. It is a small but distinctive quality of the lab.
Research foundation
Audiovisual prosody in interaction
The research basis starts with audiovisual prosody in interaction: how people use voice, facial expression and timing to manage spoken interaction. Those ideas are translated into practical agent concepts, especially Eva's focus on feedback, turn-taking, repair, emotions and emphasis.
How the lab is built
You are part of the thinking, not just the input to it
The lab leans on distributed cognition: the idea that thinking is not confined to one mind but happens across a person and the tools, artefacts and other people around them. That shows up at every layer here, not only inside a single agent. Your transcript gets reasoned over, a live analysis panel reflects the conversation back to you, and agents hand partial results to each other or back to you. The Idea Sharper pair and the Multi-Agent Solution Designer make this visible as two agent roles splitting one reasoning process, but the same principle runs through the whole site: more on distributed cognition.