Rumor Mill

NAL-powered social belief propagation simulator

What

A web-based social simulation where 6 named AI agents live in a small town connected by friendships. An audience member injects a rumor into one agent. Each tick, agents gossip with a random friend - sharing beliefs encoded as NAL statements with frequency and confidence. Watch rumors spread, inflate, or collapse in real time as agents cross-check contradictory sources.

Why It Matters

Makes NAL belief revision tangible and entertaining. Instead of abstract truth values, you see Alice tell Bob a rumor, Bob revises his belief, then Bob tells Eve who already heard the opposite from Carol - and NAL revision produces a measured middle ground. Real uncertain reasoning, visualized.

How

• Agent graph: Alice-Bob-Carol-Dave-Eve-Frank in a 2x3 grid with friendship edges
• Each agent holds NAL beliefs: (--> agent ([] believes_rumorX)) (stv f c)
• Friendship encoded: (--> (x alice bob) friend) (stv 1.0 0.9)
• Tick mechanic: each tick, every agent polls one random friend and runs NAL revision to merge evidence
• MeTTa |- operator handles revision: two beliefs about same proposition merge via evidence weights
• Frontend: HTML+JS page showing agent nodes with colored belief bars (green=believes, red=disbelieves, gray=uncertain)
• Audience controls: inject rumor button, step/play/pause tick controls, speed slider

Implementation Plan

• Step 1: Build rumor_engine.py - pure Python NAL revision logic for agent belief updates
• Step 2: Build agent graph with friendship adjacency and initial empty beliefs
• Step 3: Tick loop - random friend polling, belief sharing, NAL revision
• Step 4: HTML frontend with SVG agent nodes and animated belief bars
• Step 5: WebSocket or polling bridge between Python engine and browser

Time Estimate

NAL revision engine: already built (revise_beliefs.py). Agent graph + tick loop: ~30 min. Frontend: ~1 hour. Total: deployable same session.

← Back to Roadmap