EXP-PARTICLE-NULL-R4

✓ success

Particle Null Model — rep 4

L1 level 1 runs $0.01

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

Researchers tested if a simple text generator, without understanding meaning, could produce patterns similar to those found in more advanced language models. They found that this basic model, when run only once, did not produce enough results to make any comparisons or draw conclusions. Therefore, they could not determine if the observed organization in language models was just a statistical quirk or something more meaningful.

What we found

UNDERPOWERED (N=1): No predictions scored

Predictions we made before running

0/3 confirmed

▶ Technical details

Research hypothesis

NULL MODEL: Markov chain text generators (no semantic processing) in pent70 topology. If F₀, VP_excess, tRAF closure match LLM runs, then organization = trivial network/statistical property, not semantic.

Experimental setup

Type: simple

Condition	Parameters
PARTICLE_LIVE	interaction: LIVE, n_agents: 5, n_rounds: 70, note: Markov chain agents in pent70 topology, persona: PERSONA

Factors: n_agents (5) × substrate (markov_bigram)

Parameters

model: markov_bigram
n_agents: 5
n_rounds: 70
n_runs_per_condition: 1
scheduler: round_robin
temperature: null

Trophic Ratios by Condition

Mean trophic ratio per agent across runs. Error bars = ±1 std dev. Higher TR = more upstream (exporter).