Social dynamics
A memory of who collaborates well with whom. Productive pairings strengthen, idle pairings decay, and an explicit exploration rate prevents the firm from settling into echo chambers.
Social dynamics is the firm's measure of collaboration quality. It tracks which Participants — human or agent — produce good mission outcomes together, and uses that signal to bias future pairings.
The mechanism is borrowed from neuroscience: cells that fire together wire together. Pairings that contribute to convergent, low-Prediction-Error missions strengthen. Pairings that go idle decay. The brain analogy is deliberate — what fires together should become easier to recall.
Group composition has outsized effect on group output. Human organisations know this and pay for it with org charts, team rituals, and tenure. An agentic firm has none of those for free — every mission could draw any agent, randomly, and would.
Social dynamics gives the firm a memory of its own social fabric. Without it, every mission starts from scratch on who should be in the room. With it — and with an exploration rate guarding against complacency — the firm learns its own social geometry over time.
When two Participants are part of a mission that converges well, the edge between them grows. Strength is a function of outcome quality, not chat volume.
Edges fade when not exercised. The firm does not freeze its early pairings forever; the social map stays current with what is actually being delivered.
A configurable fraction of pairings is drawn at random, ignoring history. Without this, strong edges self-reinforce and the firm collapses into a clique.
Beyond the quantitative signal, the firm tracks discursive behaviour: building on a peer's point, challenging a causal gap, peer questions. Convergence is not allowed without at least one genuine challenge.