Language-model social dynamics

LLM Social Simulation

Use LLM social simulation to model how language-model actors with roles, memory, incentives, and context may react to the same scenario over multiple rounds.

Create an LLM social simulation See how it works

Built for scenario planning, not certainty. Every output should be reviewed against real evidence before an operating decision.

Scenario tape / illustrative

LLM social simulation for a public reaction scenario

Social rounds

Seed audienceFirst interpretation forms

Influence clusterNarrative frame spreads

Opposing groupCounter-reaction appears

Dominant force

Social interpretation

Amplifier

Memory and framing

Evidence gap

Group trigger

Input

Scenario brief, actors, context, evidence

Engine

LLM actors + social reaction rounds

Output

Narrative paths and group pressure risks

What to bring

Bring the social situation you want to stress-test.

LLM social simulation works best when the source material defines the event, the groups involved, the context they remember, and the evidence that shapes how they interpret the scenario.

Scenario and source material

The incident, launch, policy, story event, market move, or public question you want simulated with enough evidence to ground reactions.

Social actors and roles

Audience segments, communities, institutions, critics, supporters, buyers, employees, media voices, or other groups with distinct motives.

Memory and context

Prior events, public claims, history, unresolved tensions, repeated narratives, constraints, and facts that actors should remember.

Where it earns its keep

See how group interpretation can evolve before it becomes visible.

Social reaction rarely moves in one clean line. MiroFish works as an LLM social simulation tool for inspecting how roles, memory, and influence can turn one event into competing paths.

Public reaction01

Model how audiences interpret the same evidence differently.

Explore how supporters, skeptics, media accounts, affected groups, and institutions may form distinct narratives.

Community dynamics02

Watch reaction move through groups over rounds.

Simulate how one group's response can trigger amplification, fatigue, conflict, or reframing in another group.

Narrative pressure03

Find the memory that changes the next reaction.

Identify which prior claim, unresolved issue, or context detail makes the scenario more fragile.

LLM social simulation needs roles, memory, and interaction.

MiroFish turns source material into language-model actors with motives, context, constraints, and memory, then runs social reaction rounds so the path can be reviewed.

Stage 1
Ground the scenario
Upload source material and define the decision, event, or message you want to test.
Stage 2
Build the actor graph
Map the people, groups, incentives, constraints, and memory that shape the reaction.
Stage 3
Run reaction rounds
Let the simulated actors respond over multiple rounds so the second-order path appears.
Stage 4
Question the report
Review the trajectory, risks, weak assumptions, and what evidence would change the conclusion.

What the report should answer

The report should make social dynamics inspectable.

Which group forms the first dominant interpretation
Which memory or claim amplifies the reaction
Which actor cluster creates a counter-frame
Where the scenario depends on weak or missing evidence
Which follow-up question should be tested next

Boundary conditions

Not a substitute for real social data.

Not guaranteed prediction of real human behavior
Not polling, live social listening, or demographic measurement
Not useful without source material, roles, or relevant context
Not a replacement for expert review in high-stakes public decisions

Why MiroFish

LLM social simulation should reveal interaction, not just sentiment.

Generic AI can summarize likely sentiment. MiroFish models social actors, memory, narrative pressure, and multi-round reactions.

Social actors

Generic AI

Collapsed into one audience

MiroFish

Distinct roles with motives, memory, and context

Reaction path

Generic AI

One-shot sentiment summary

MiroFish

Multi-round social interpretation and counter-reaction

Review

Generic AI

Advice without inspectable structure

MiroFish

Actor graph, weak assumptions, and evidence gaps

FAQ

LLM social simulation, grounded in scenario evidence.

Use it to explore how language-model actors can represent bounded social reactions without pretending to measure reality directly.

What is LLM social simulation?+

LLM social simulation uses language models as bounded actors inside a shared scenario so teams can inspect how group reaction, interpretation, and narrative pressure may evolve.

How is it different from multi-agent simulation AI?+

Multi-agent simulation is the broader method of modeling interacting agents. LLM social simulation focuses specifically on social interpretation, group reaction, roles, memory, and narrative dynamics using language-model actors.

Can it predict real public opinion?+

No. It is a scenario planning tool, not a polling system. It helps surface plausible reaction paths, weak assumptions, and missing evidence before decisions are made.

What should I upload?+

Upload scenario briefs, articles, public comments, policy drafts, launch materials, research notes, community context, incident timelines, or other evidence that defines the social situation.

Who should use it?+

Researchers, founders, policy teams, communications teams, product marketers, and operators can use it when social interpretation may change the outcome of a decision.

Related simulation paths