← back to timeline

EXP-TOPO-STAR

artifact

Star Topology — Does Structural Hub Create Hierarchy?

2026-03-14 L3 level paper3 1 runs $0.00
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

Three AI agents communicate in a text space. One of them is the "hub": sees both others. The other two are "spokes": see only the hub, not each other. Will the hub become a word "source" — the top of hierarchy? Or does structural advantage not guarantee dominance?

What we found

UNDERPOWERED (N=1): No predictions scored

Predictions we made before running

0/5 confirmed

Technical details

Research hypothesis

OPEN QUESTION: "degree distribution operationalization: need SPECIFIC threshold — density < X% → F₀_excess > Y AND degree heterogeneity > Z → hierarchy. without this, mechanism = unfalsifiable."

CONTEXT: Cycle topology (exp-topo-cycle, N=1) KILLED hierarchy (TR 0.071 vs 0.444 baseline). Cycle = uniform degree (in=1, out=1 for all). Complete = uniform but higher (in=2, out=2). Both uniform degree topologies produce different hierarchy levels.

KEY QUESTION: Does ASYMMETRIC degree create hierarchy? Star topology introduces a HUB with higher in-degree than spokes.

  • Hub (Alpha): reads Beta AND Gamma (in-degree=2)
  • Spoke Beta: reads only Alpha (in-degree=1)
  • Spoke Gamma: reads only Alpha (in-degree=1)

If hub becomes source → degree heterogeneity CAUSES hierarchy. If hub doesn't → hierarchy in complete topology is NOT from degree alone.

PREDICTIONS grounded in:

  • Cycle result: uniform degree → flat (TR=0.071)
  • Complete baseline: uniform-but-high degree → hierarchy (TR=0.444)
  • Theory: hub has informational advantage (sees 2 streams), spokes are isolated from each other (convergence must transit through hub)

BASELINE DATA: Complete topology (exp-asym PERSONA_LIVE, N=9): TR_range = 0.444 ± 0.15 Cycle topology (exp-topo-cycle, N=1): TR_range = 0.071 Full decay complete (exp-forget, N=1): TR_range = 0.015

SECOND in designed-topology series: cycle (done) → star (this) → chain → complete (baseline).

Experimental setup

Type: simple

Condition Parameters
STAR_LIVE topology: STAR, interaction: LIVE, note: Hub (Alpha) reads Beta+Gamma. Spokes (Beta, Gamma) read only Alpha.

Factors: topology (STAR)

Parameters

n_agents
3
n_rounds
48
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).