EXP-CA-RULE-MULTIGEN-5GEN

… pending

CA + multi-gen 5-generation control — resolves step-3 depth confound (paper6 step 3 control)

2026-05-04 L6 level paper6 4 runs $0.06

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

Researchers are investigating why a program's complexity decreased when they used fewer generations in their experiment. They ran the experiment again with the same number of generations as before, but with a different underlying system. This will help them determine if the decrease in complexity was due to fewer generations or a different kind of competition within the system.

▶ Technical details

Research hypothesis

Step 3 variance (N=10) showed mean depth 3.9 (vs step-2's 7.1). Confound: step-3 used 3 generations vs step-2's 5. Cannot distinguish whether depth drop is from (a) fewer generations or (b) substrate competition for proposal bandwidth. This control: matched 5 generations × 20 rounds (same as step-2). Same engine with CA substrate. If depth ~7 returns → drop was generation-count. If depth stays ~3-4 → CA competition genuinely shifts the program. Pre-registered: - GENERATION-COUNT hypothesis: depth ~6-8 (matches step-2 range) - SUBSTRATE-COMPETITION hypothesis: depth ~3-5 (matches step-3 N=12) - Either result is informative. Also tracking: do CA-axis acceptance share and reasoning grounding stay ~70% / ~100% with longer runs?

Experimental setup

Type: simple

Condition	Parameters
COUPLED_5GEN	coupling: ALL, n_runs: 4

Factors: coupling (all)

Parameters

n_agents: 4
n_rounds: 100
model: google/gemini-2.5-flash-lite
temperature: 0.85

autopoiesis paper6-step3-control generation-count-confound