EXP-COMPETENCY-N5-ROT2

informative

N=5 Competency — rotation=2 (Gamma-first)

2026-04-02 L3 level paper3 1 runs $0.01

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

We rotate who speaks first in a 5-agent conversation. Reanalysis of 13 runs shows N=3 position dominance was an artifact — at N=5, 70% of variance is noise. Adding a third rotation to fill the position×persona matrix and resolve whether any systematic structure exists.

▶ Technical details

Research hypothesis

POSITION vs PERSONA DECOMPOSITION at N=5 with rotation=2.

CONTEXT: N=3 competency-rotation showed POSITION η²=0.600 >> persona η²=0.246. N=5 reanalysis (12 pent70 + 1 revord) shows BOTH effects collapse: Full: η²_pos=0.162, η²_per=0.134, residual=70.4% Middle: η²_pos=0.139, η²_per=0.111, residual=74.9% Kendall W=0.214 (weak rank consistency)

PROBLEM: Only 1 reversed-order run. Position×persona cell table is sparse. rotation=2 gives: Gamma(pos1), Delta(pos2), Epsilon(pos3), Alpha(pos4), Beta(pos5) This puts Alpha at pos4 (was pos1 in pent70), giving new cells.

KEY TEST: Does Alpha maintain high TR at pos4? If so → persona. If drops → position. Similarly: Gamma at pos1 — does it become highest? If so → position > persona.

Combined with pent70(rot=0) and revord(reversed), gives 3 position mappings.

Experimental setup

Type: simple

Condition	Parameters
ROT2_LIVE	interaction: LIVE, n_agents: 5, n_rounds: 70, rotation: 2, note: N=5, 70 rounds, rotation=2. Gamma speaks first.

Factors: n_agents (5) × rotation (2)

Parameters

n_agents: 5
n_rounds: 70
n_runs_per_condition: 1
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).