EXP-RNS-G2

informative

Ratchet Null (single-agent null model) — G2

2026-03-27 L6 level paper3 1 runs $0.01

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

This experiment tested a simplified scenario with just one artificial agent to see if it could learn any useful strategies on its own. However, the experiment was not set up to gather enough information to draw any conclusions. Therefore, no predictions could be made about what the agent might learn.

What we found

UNDERPOWERED (N=1): No predictions scored

Predictions we made before running

0/1 confirmed

▶ Technical details

Research hypothesis

G2 of single-agent null model. 1 agent(s) × 30 rounds.

Experimental setup

Type: single_condition

Condition	Parameters
G2_NULL	persona: PERSONA, interaction: SOLO

Factors: persona (PERSONA) × interaction (SOLO)

Parameters

n_agents: 1
n_rounds: 30
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).