← back to timeline

EXP-N3-BALANCED-ROT

informative

N=3 Balanced Rotation Latin Square — Interaction Term at Small N

2026-04-03 L3 level paper3 9 runs $0.04
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

Three AI agents with different thinking styles talk in a shared text space. We rotate who speaks first to separate the effect of personality from the effect of speaking order. At group size 5, personality-position COMBINATIONS matter (interaction effect). Does this interaction vanish at group size 3, where speaking order dominates? If yes, crossing from 3 to 5 agents is a threshold — like switching from rigid hierarchy to flexible self-organization.

Technical details

Research hypothesis

N=3 BALANCED ROTATION: η²(interaction) at Small N

CONTEXT: exp-balanced-rot (N=5, 15 runs): η²(pos)=0.007, η²(per)=0.146, η²(int)=0.082. N=3 competency-rotation: η²(pos)=0.60, η²(per)=0.25. BUT confounded — no balanced design, so interaction term not cleanly estimable.

Dochkina 2603.28990: capability threshold — below capacity, structure (position) dominates; above capacity, hybrid interaction wins.

HYPOTHESIS: At N=3, η²(interaction) ≈ 0, because positional constraint is so strong (3 slots, no slack) that persona cannot modulate positional effects. The N=3→N=5 transition IS the "capability threshold crossing."

DESIGN: Latin square: 3 cyclic rotations × 3 seed sets = 9 runs. Each persona appears in each position EXACTLY ONCE per seed set. Rot0: [Alpha, Beta, Gamma] Rot1: [Beta, Gamma, Alpha] Rot2: [Gamma, Alpha, Beta]

r(pos, persona_id) = 0.00 by construction — ORTHOGONAL design. Enables clean 2-way ANOVA: position × persona → η² decomposition. Directly comparable to exp-balanced-rot (N=5) methodology.

KEY PREDICTION: η²(int) < 0.03 at N=3 (vs 0.082 at N=5). If confirmed: N-transition = Dochkina's capability threshold.

DECISIVE TEST from NOW.md #1: "If near-zero → confirms N-transition = Dochkina's capability threshold. Zero-cost, decisive test." — except this needs real data since we don't have balanced N=3 yet.

Experimental setup

Type: factorial

Condition Parameters
LIVE interaction: LIVE, n_agents: 3, n_rounds: 48, note: Standard N=3 live interaction. Latin square rotations via run_matrix.

Factors: rotation (0, 1, 2) × seed_set (S1, S2, S3)

Parameters

n_agents
3
n_rounds
48
n_runs_per_condition
9
model
gemini-2.5-flash-lite
temperature
0.9
scheduler
round_robin

Trophic Ratios by Condition

Mean trophic ratio per agent across runs. Error bars = ±1 std dev. Higher TR = more upstream (exporter).