Multi-agent simulation, in plain terms
A multi-agent simulation models a group as many separate agents, each with its own goals, rather than as one average. When the agents interact, patterns emerge that you cannot get from a single forecast — which is exactly what makes reactions predictable.
Why emergence matters
Real crowds are not uniform. One loud account can flip a segment; a quiet majority can absorb a change without a ripple. Because each MiroFish agent acts on its own persona and motives, effects like these appear on their own instead of being assumed in advance.
From agents to a report
MiroFish extracts the actors from your seed into a knowledge graph, runs them as agents across rounds on a simulated social surface, and then condenses what happened into a readable prediction report.
How many agents run?
It varies by scenario and run size; the point is enough diversity of persona for group dynamics to emerge rather than a fixed headcount.
Is this the same as agent-based modelling?
It is the same family of idea, applied to opinion and narrative dynamics and driven by language-model agents.