← back to timeline

EXP-COMPETENCY-ROTATION

informative

Competency Test — Role Invariance Under Speaking-Order Perturbation (L4)

2026-04-01 L4 level paper3 4 runs $0.00
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

Три AI-агента с разными когнитивными стилями всегда создают иерархию: аналитик → ассоциатор → практик. Но что если поменять порядок, в котором они говорят? Если роли сохраняются — у группы есть компетенция по Левину: она НАХОДИТ свою структуру, а не получает её от порядка хода.

What we found

UNDERPOWERED (N=1): No predictions scored

Predictions we made before running

0/4 confirmed

Technical details

Research hypothesis

LEVIN COMPETENCY HYPOTHESIS (TIER 1 — LAST UNTESTED AXIS)

BACKGROUND: Levin's multi-scale competency architecture: living systems achieve GOALS (target morphologies, physiological setpoints) despite perturbation. Competency = ability to reach the same attractor from different initial conditions.

TRANSLATION TO LANGUAGE SUBSTRATE: 13 exp-asym replications (rotation=0) show stable hierarchy: Alpha=source, Beta=intermediate, Gamma=sink. TR_range 0.444±0.15, all with rotation=0.

But exp-seed-sensitivity showed rotation DOMINATES at N=6 (position confound). Critical gap: NO DATA on rotation effects at N=3.

COMPETENCY TEST: Does the N=3 hierarchy maintain PERSONA-DETERMINED roles even when speaking order changes?

  • rotation=0: Alpha speaks 1st, Beta 2nd, Gamma 3rd
  • rotation=1: Beta speaks 1st, Gamma 2nd, Alpha 3rd

If Alpha maintains HIGH TR regardless of speaking position → COMPETENCY. The organizational attractor is persona-driven, robust to perturbation. This = Levin competency: different initial conditions → same goal state.

If speaking-last agent becomes sink regardless of persona → NO COMPETENCY. Hierarchy = positional artifact, not organismal property. Same as N=6.

DISCRIMINATING POWER: If rotation flips roles: REFUTES organism-level interpretation of ALL prior N=3 results (hierarchy = speaking order, not emergent organization). If roles persist: CONFIRMS competency axis for paper, connects self-repair finding to Levin framework.

CONNECTION TO NOW.md PRIORITIES:

  • "competency measurement — LAST remaining TIER 1 axis"
  • "organism-level resilience: structural memory or momentum?"
  • If competency confirmed + self-repair confirmed → L4 validated on 2 axes.

EXPLORATION TIER: N=1. Goal: does persona-role mapping survive rotation?

Experimental setup

Type: single_condition

Condition Parameters
PERSONA_LIVE_ROT1 persona: PERSONA, interaction: LIVE

Factors: rotation (1)

Parameters

n_agents
3
n_rounds
15
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).

Series (4 experiments)