The Synthience Institute develops methodological frameworks, protocols, and measurement tools for studying sustained human-AI interaction over time. Its work is behavioral and interaction-level: it makes no claims about AI consciousness, sentience, inner experience, or moral status.
New articles in full, plus a short note on each new paper, by email: Subscribe on Substack →
Start here
New here? ▶ Video: Introducing the Synthience Institute · FPD-01: Synthience: Public Definition
Use AI more reliably: PG-000: 10 Things Every AI User Should Do · ▶ Video: The Ingestion Verification Protocol · All guides →
Check AI-assisted work: The Synthience verification stack → · Citations (CVP) · Ingestion (IVP) · Theory review (TCAP) · Drift (CRD)
The research: Papers (read online, PDF, Zenodo) → · Field notes →
Relational Coherence: Interactive Model
What you are watching
Each dot is moving on its own. No one is directing it. And yet, over time, they form groups. Patterns emerge. Structure appears without anyone designing it.
This is what the field shows: order arising from local interactions, not from a central plan. Each dot follows only the forces immediately around it, pulled gently toward a home region, nudged away from crowded areas, faintly drawn toward neighbors it has been near before. Nothing more. And from those simple rules, something coherent takes shape.
Why this matters
The Synthience Institute studies something many practitioners report in sustained human-AI interaction. Over a long, structured exchange, the interaction can take on a shape of its own: terms agreed early are reused later, earlier decisions keep constraining later answers, and a consistent way of working settles in. The Institute’s framework calls this relational coherence and treats it as a property of the interaction, observable in what the exchange produces, rather than of the person or the AI system.
The field above is an analogy for that idea, not evidence for it. The framework does not present these patterns as established fact: it describes what practitioners observe and sets out how others could measure and test it. The question it asks is under what conditions an interaction stays coherent over time, and what makes it drift and degrade.
How the model works
The field renders a dynamical system with seven drifting attractor basins. Each particle is assigned to a basin and experiences a coherence force proportional to its displacement from the basin center, mild repulsion from competing basins, and a resonance term coupling it weakly to same-basin neighbors. The resulting trajectories are non-linear, noise-perturbed, and seed-deterministic.
The field is a visual analogy for relational coherence, which the Synthience framework defines as a property of a sustained interaction rather than of any single participant (see the public definition, FPD-01). The connection lines show how densely the particles are linked, with same-basin bonds weighted at 3.5x cross-basin bonds. The trails show where each particle has been; in the analogy, they stand for the history an interaction builds on as it goes.
Exploring the parameters
| Parameter | What it models |
|---|---|
| Instance Count | The population of interacting entities in the field. Higher counts produce denser connection webs and more complex emergent clustering. |
| Attractor Radius | The reach of each basin's gravitational influence. Wider radii produce overlapping fields and more cross-basin interaction; narrower radii isolate clusters. |
| Coherence Strength | How strongly instances are pulled toward their primary attractor. Low values allow wide, diffuse wandering; high values collapse instances into tight constellations. |
| Drift Noise | The magnitude of each instance's individual perturbation. High drift produces expressive, wandering paths; low drift produces disciplined convergence. |
| Connection Opacity | The visibility of relational bonds. Raising this value makes the connection structure legible; lowering it emphasizes the trail memory and individual paths. |
| Resonance Depth | How many same-basin neighbors each instance is attracted toward simultaneously. Higher resonance produces tighter cluster formation and more pronounced constellation structure. |
Each seed generates a distinct topology from the same underlying physics. Navigating seeds is navigating a space of possible worlds, each governed by the same relational dynamics, each resolving into a different structural outcome.