EXP-BROADCAST
✓ successBroadcast Topology — In-Degree=0 Source vs Star In-Degree=2 Sink
In plain language
Three AI agents communicate, but only one — Alpha — generates words that others can read. Beta and Gamma read Alpha but cannot see each other and nobody reads them. Alpha is a "broadcaster": pure source, zero input. Will this create the strongest hierarchy — or will isolation from feedback make Alpha drift into irrelevance?
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."
KEY FINDING FROM EXP-DIV (2026-03-22): Concentrated imbalance (few nodes highly asymmetric) → COHERENT hierarchy (low F₀). Distributed imbalance (all nodes mildly asymmetric) → INCOHERENT hierarchy (high F₀). Cohen's d = -4.31 on hub asymmetry direction (REVERSED from prediction).
KEY FINDING FROM STAR TOPOLOGY: Hub (Alpha, in-degree=2) became SINK, not source. TR=0.409 (LOWEST). Mechanism: high in-degree → vocabulary absorption → imports > exports.
THIS EXPERIMENT: BROADCAST topology — the STRUCTURAL INVERSE of star.
- Alpha: in-degree=0 (reads NOBODY), out-degree=2 (both Beta and Gamma read Alpha)
- Beta: in-degree=1 (reads Alpha only), out-degree=0 (nobody reads Beta)
- Gamma: in-degree=1 (reads Alpha only), out-degree=0 (nobody reads Gamma)
PREDICTION grounded in star lesson + DIV paradox: If star hub (in=2, out=2) is SINK because of in-degree, then broadcast Alpha (in=0, out=2) should be pure SOURCE. Alpha's in=0 means literally cannot absorb. Must generate original vocabulary. Beta and Gamma are vocabulary RECEIVERS with no audience → dead-end sinks.
CRITICAL COMPARISON: broadcast vs star vs chain | | Alpha in | Alpha out | Beta/Gamma in | Beta/Gamma out | |----------|----------|-----------|---------------|----------------| | Star | 2 | 2 | 1 | 1 | | Chain | 0 | 1 | 1 | 0-1 | | Broadcast| 0 | 2 | 1 | 0 | | Complete | 2 | 2 | 2 | 2 | (only Beta reads Alpha; both read Alpha)
Broadcast vs Chain: both have Alpha in=0, but broadcast has Alpha out=2 (two sinks). If broadcast shows STRONGER hierarchy than chain → out-degree amplifies source effect. If same → in-degree alone determines source/sink, out-degree irrelevant.
Broadcast vs Star: MAXIMALLY INVERTED structure. Star Alpha: max in, max out → SINK. Broadcast Alpha: zero in, max out → should be SOURCE. If both show Alpha as SAME role → in-degree doesn't cause sink (star lesson wrong).
BASELINE DATA: Complete (exp-asym, N=9): TR_range=0.444±0.15, VP=0.075, C=0.344 Star (exp-topo-star, N=1): TR_range=0.141, VP=0.243, C=0.778, Alpha=SINK Chain (exp-topo-chain, N=1): TR_range=?, VP=?, Alpha=? Cycle (exp-topo-cycle, N=1): TR_range=0.071, VP=0.167, C=0.715
FOURTH in designed-topology series: cycle → star → chain → BROADCAST (this). Completes the 2×2: {symmetric, asymmetric} × {cyclic, acyclic} topology space.
Experimental setup
Type: simple
| Condition | Parameters |
|---|---|
| BROADCAST_LIVE | topology: BROADCAST, interaction: LIVE, note: Alpha reads no agents (in=0). Beta reads only Alpha (in=1). Gamma reads only Alpha (in=1). Nobody reads Beta or Gamma. |
Factors: topology (BROADCAST)
Parameters
- n_agents
- 3
- n_rounds
- 48
- n_runs_per_condition
- 1
- model
- gemini-2.5-flash-lite
- temperature
- 0.9
- scheduler
- round_robin