Privacy & values
Agent Interaction Lab explores communication-aware agents, multi-agent solution design and creative prototyping. The public demos are intentionally lightweight and designed around explicit user action.
The mission behind this lab is to explore how AI can make people's lives easier and better, and open up opportunities that were not there before. The focus is on AI as a co-creation partner: through the interaction between people and agents, and between agents themselves, new collaboration structures can emerge that create value no single party could have created alone.
This is an open exploration rather than a fixed destination with a known answer in advance. The aim is to discover, step by step, what a value-creation cycle around human-agent collaboration could look like, and what becomes possible that was not possible before.
The demos process input only after a visitor starts an interaction, submits a form or requests an expert review. No unnecessary personal data storage is part of the current public prototype.
Conversation text is sent to OpenAI's API (based in the United States) to generate agent replies, and is not retained by OpenAI beyond generating that reply. This site is used worldwide and this transfer relies on OpenAI's standard contractual safeguards.
If you switch its speech mode to "Human," the Emotion Regulator agent speaks with an AI-generated voice from Sesame instead of your browser's built-in voice, used only to generate that agent's spoken replies, never to process anything you say.
Generated results are sent for human-in-the-loop review only when the visitor chooses that route. The same applies when someone explicitly allows a result to help improve the agent prompts.
The agents can support reflection, communication practice and prototype thinking, but they do not provide diagnosis, therapy, medical advice or clinical assessment.
The current public setup does not use advertising cookies or third-party trackers. Some first-party, aggregated technical data is kept to keep the demos reliable and fair to use for everyone.
Every agent here is an AI system, not a person, built on third-party language models with our own prompts and safety rules on top. Answers can be wrong, and none of the agents provide diagnosis, therapy or medical advice.
Some agents can ask permission to use your tone and pace of speaking to respond better. This is always asked directly in the conversation: you choose yes or no, and analysis only starts if you say yes. Nothing about your voice is recorded or stored, only short, aggregated labels are used for that reply.
Agents adjust language, tone and explanation depth within clear limits. A dedicated page describes each adaptation, how long it lasts and when consent is needed.
See how agents adapt