← back to timeline

EXP-STAR-N5

artifact

Star N=5 — Does Hub-as-Sink Scale with Fan-In?

2026-03-15 L3 level paper3 1 runs $0.01
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

Five AI agents in a star topology: one "hub" sees all four others, the rest see only the hub. With three agents, the hub became a word "absorber." What happens with five? Will the effect intensify or will something qualitatively new emerge?

What we found

UNDERPOWERED (N=1): No predictions scored

Predictions we made before running

0/5 confirmed

Technical details

Research hypothesis

OPEN QUESTION: "emergence threshold: at what N agents + interaction density do emergent properties appear?"

CONTEXT: Star topology at N=3 (exp-topo-star) confirmed hub-as-SINK:

  • Alpha (hub, in-degree=2): TR=0.409 (lowest), E/I=0.69, Red=1.24 (highest)
  • Hub absorbs redundant info from both spokes
  • VP=0.243 (highest of all topologies — hub amplifies convergence)

KEY QUESTION: Does increasing fan-in (N=5, hub in-degree=4) AMPLIFY these effects? Three competing hypotheses: (A) LINEAR scaling: hub-as-sink deepens proportionally (4 input sources → ~2× redundancy) (B) THRESHOLD/SATURATION: 4 inputs overwhelm hub → convergence collapses or hierarchy inverts (hub can't absorb, starts generating from overload) (C) EMERGENT: new dynamics appear at N=5 that don't exist at N=3 — spoke cliques, intermediate hierarchy levels, hub fragmentation

BASELINE DATA: Star N=3 (exp-topo-star, N=1): TR_range=0.141, VP=0.243, C=0.778, RP=0.072 Complete N=3 (exp-asym, N=9): TR_range=0.444, VP=0.075, C=0.344, RP=0.147 Complete N=5 (exp-pent, N=1): available for comparison

GROUNDING: Gershenson E/S/C framework — complexity (C) at intermediate coupling. Star N=5 increases coupling for hub only. Does C peak or plateau?

Experimental setup

Type: simple

Condition Parameters
STAR_N5_LIVE topology: STAR_N5, interaction: LIVE, note: Hub (Alpha) reads Beta+Gamma+Delta+Epsilon. All spokes read only Alpha.

Factors: topology (STAR_N5)

Parameters

n_agents
5
n_rounds
50
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).