Aligned Failure

A formal toy model of modal conflict: each agent's marginal reliability can increase while the reliability of the community majority decreases, because failures become aligned.
ModalConflictModel0_2

1. Summary

Regime Agent analytic Majority analytic Agent empirical Majority empirical
Independent majority reliability:
P(M correct) = Σk=tn C(n,k)pk(1-p)n-k =
Hover over the summation, a distribution row, or a grid cell. The distribution table highlights the corresponding columns in the grids.

2. Distribution over number of correct agents

K is the number of correct agents in a run/world. The majority is correct when K ≥ t. Independent evidence spreads probability across many rows; a single source puts nearly all analytic mass at the extremes. Each mini-bar is scaled from 0 to 1.
K Majority? Independent analytic Independent empirical Single-source analytic Single-source empirical

3. Modal grids

correct incorrect M = community majority
The canvases always fit the panel width. With many runs, each column becomes thinner rather than overflowing. Switch to raw run order to see that the sample is generated randomly.
Independent evidence: dispersed failures
errors partly cancel
Single source: aligned failures
errors align