Harmonia

A protocol for structured human-AI deliberation

What is it?

Harmonia is a formal protocol that structures dialogue between humans and AI systems through explicit reasoning cycles. Instead of free-form chat, every exchange follows a Perceive → Synthesize → Model → Reflect loop that makes AI reasoning transparent and auditable.

Core idea: AI should show its reasoning, not just its conclusions.

How is it different?

Current AI assistants (ChatGPT, Claude, Gemini) operate as black boxes:

You ask → they answer
No visibility into reasoning process
No structured way to challenge assumptions
Trust is implicit, not earned

Harmonia makes reasoning explicit:

Every response declares its sources and assumptions
Uncertainty is quantified, not hidden
Users can inspect and challenge any step
Trust is built through transparency

The PSMR Cycle

Each deliberation turn follows four phases:

Perceive — What information is available? What context matters?
Synthesize — How do these elements connect? What patterns emerge?
Model — What hypotheses explain this? What predictions follow?
Reflect — What confidence level? What assumptions? What's missing?

Where to start

📦 harmonia-kernel 📖 Open Trust Infrastructure

The kernel contains the Python reference implementation.
The book provides theoretical foundations (14 chapters, PDF available).

Canonical reference

Berlocher, I. (2025). Open Trust Infrastructure: Language, Cognition, and Trust in Agentic Systems. Zenodo.
DOI: 10.5281/zenodo.17940484