Bandit networks & transient diversity (Zollman 2007, 2010), his assumptions only. Old theory pays at exactly 0.5; the new one at 0.5 + ε. Will the community find out? · ← tutorial · the LLM extension →
Presets
Community
who sees whose results — Zollman's headline: sparser networks are slower but likelier to find the truth
community size — bigger communities are harder to tip into premature consensus
The problem
how much better the new theory really is (its payoff rate is 0.5 + ε) — small ε = a hard problem, easy to get wrong
how much data each agent generates per round — the size of each "study" before everyone updates
The scientists
Zollman's "extreme beliefs": how much contrary evidence it takes to move an agent off its starting opinion
Simulation
how many independent communities to simulate — more runs, steadier percentages
the horizon: runs still divided at this point count as "unsettled" (permanent diversity)
random seed — same seed + same settings reproduces the exact same result