← back to timeline

EXP-PERSONA-SWAP

artifact

Persona vs Position — Does Generalizer Persona Drive Trophic Advantage?

2026-03-10 L2 level paper3 4 runs $0.00
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

Agent Zeta (the "generalizer") consistently occupies the top of hierarchy. But why — because of the thinking type (generalization creates vocabulary that others adopt) or because of the name/position? We swap the roles of Zeta and Alpha: if the generalizer is on top again — it is the role, not the name.

What we found

No predictions scored

Predictions we made before running

0/5 confirmed

Technical details

Research hypothesis

QUESTION: In exp-seed-sensitivity, Zeta (generalizer persona) was consistently highest-TR agent at N=6 (mean 0.611, SD 0.093, vs others ~0.47). Meanwhile exp-may-sigma showed Alpha (position 1) dominates at N=7.

Two competing explanations: (A) PERSONA EFFECT: The generalizer persona ("abstract and generalize, extract rules, name patterns, propose frameworks") produces vocabulary that others adopt — inherently top-of-hierarchy regardless of agent name or position. (B) NAME/POSITION EFFECT: Something about being "Zeta" (6th in order, name recognition, or round-robin position) creates the advantage.

TEST: Swap Alpha and Zeta's personas at N=6. Alpha gets generalizer persona, Zeta gets analytical persona. Everything else identical to exp-seed-sensitivity.

Prior evidence:

  • exp-seed-sensitivity: Zeta mean TR=0.611 (N=16 runs), others ~0.47
  • exp-may-sigma: Alpha (position 1) highest at N=7 low-σ, Zeta lowest
  • Position ANOVA at N=6: F=0.98, p=0.443 (not significant for TR_range)
  • But per-agent analysis shows Zeta consistently top — this is DIFFERENT from rotation/position driving TR_range variance

Experimental setup

Type: simple

Condition Parameters
SWAP persona: SWAPPED, interaction: LIVE, n_agents: 6, n_rounds: 48, note: Alpha gets Zeta's generalizer persona, Zeta gets Alpha's analytical persona. Others unchanged.

Factors: persona_assignment (SWAPPED)

Parameters

n_agents
6
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).

Series (4 experiments)