More than building agents
I research and design how people communicate with agents: verbally, non-verbally, visually and interactively.
Human communication meets intelligent agents
Agent Interaction Lab
Where human communication and agent technology meet.
Exploring how humans communicate, collaborate and create with AI agents. I combine software engineering, psychology and research in speech and non-verbal communication to explore practical agent-based systems.
Software, psychology and speech
My unique expertise combines software engineering, psychology, speech communication, non-verbal communication and human-agent interaction.
Serious, technical, creative
The lab presents practical agent concepts that can grow into coaching tools, organisational design support and creative 3D workflows.
About
Expertise at the intersection of communication research and agent engineering.
My name is Pashiera Barkhuysen and I combine software engineering with research in communication, psychology, speech and non-verbal behaviour. During my PhD research, I studied how people communicate in spoken interaction and what happens when communication problems arise, including in interactions with agents.
This site explores the next step: AI agents that do not only answer questions, but observe, analyse, collaborate, design and help improve human-agent interaction.
Research foundation: Audiovisual prosody in interaction .
Software engineer
Psychology background
Speech interaction research
Non-verbal communication
Human-agent interaction
Creative 3D workflows
Demo 01
Eva Communication Coach
A simple showcase where a user talks to Eva, a communication avatar. The input is transcribed or typed, then analysed across three research-inspired dimensions: feedback, turn-taking, emotions and emphasis.
Feedback
Turn-taking
Emotions
Emphasis
Problem detection
Finality cues
Eva Communication Coach
How this interaction works
Eva asks a short question, listens for up to one minute, then analyses the transcript for feedback, turn-taking, emotions, emphasis and repair moments.
Ready for interaction
60s
Healthcare conversations
Practise explaining symptoms, asking clarifying questions or checking mutual understanding.
Customer contact
Train difficult service conversations before they happen with real customers.
Coaching and education
Make feedback, hesitation, emphasis and repair moments more explicit and discussable.
Accessible communication
Support people who benefit from clear, explicit conversational feedback without making diagnoses.
Demo 02
Multi-Agent Solution Designer
Two agents analyse a practical organisational challenge from different roles: human psychology, communication psychology and technical system design come together in a practical solution concept. The agents work in a turn-by-turn chain: psychology input becomes software input, and software output becomes the next psychology input. The result may include an agent, but it can also become a workflow, software prototype, training tool or decision-support system.
Demo 03
Creative 3D Designer
A creative agent that designs 3D objects from text prompts and explains how the concept could become a printable object.
Expert Review
Start with an agent-generated concept. Upgrade it with a human expert review.
The agents can give you a first direction quickly. If you want a sharper version, you can request a compact expert review: I refine the output, improve the reasoning and turn it into a more useful next-step plan.
View pricing planInstant draft
The system creates an initial communication analysis, solution concept or 3D design direction.
Optional expert layer
Request a human review when you want more clarity, feasibility and practical next steps.
Improved version
Receive a refined version that is easier to act on, share internally or turn into a prototype.
Contact
Interested in agents that take communication seriously?
Get in touch for collaboration, advice, research prototypes or technical concept development around human-agent communication, speech interaction, multi-agent systems and creative 3D workflows.