EXP-DEVELOPMENTAL-TRAVERSAL
… pendingDevelopmental Traversal — do organizational levels emerge in order from minimal agents?
In plain language
We give five AI agents the simplest possible instructions — just "respond to what you see" — and let them talk for 1000 rounds under resource pressure. Do they develop the same organizational structures that usually require detailed role descriptions? And if they build these structures from scratch, can they also REORGANIZE them — something pre-instructed agents cannot do?
▶ Technical details
Research hypothesis
DEVELOPMENTAL TRAVERSAL HYPOTHESIS. Occupation != traversal. Agents PLACED at high organizational levels cannot reorganize because they did not CLIMB through lower levels. Crystal ceiling = absence of developmental history. If minimal agents (no persona prompts, no behavioral instructions) interact for 1000 rounds under scarcity pressure, organizational levels should emerge IN ORDER (L1 before L2 before L3 etc.). Furthermore, vertical capacity (Bateson Learning II: type change, not just quantity change) should appear AFTER horizontal traversal through organizational levels — and should be ABSENT in agents PLACED at high levels without developmental history. This is the decisive experiment: if developmental history unlocks vertical capacity that placement cannot, the crystal ceiling is about HISTORY, not substrate. Grow, don't build.
Experimental setup
Type: developmental_traversal
| Condition | Parameters |
|---|---|
| DEV-07-7K | prompt_type: MINIMAL, temperature: 0.7, budget: 7000, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 10 |
| DEV-03-7K | prompt_type: MINIMAL, temperature: 0.3, budget: 7000, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 1 |
| DEV-10-7K | prompt_type: MINIMAL, temperature: 1, budget: 7000, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 1 |
| DEV-14-7K | prompt_type: MINIMAL, temperature: 1.4, budget: 7000, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 1 |
| DEV-07-UNLIM | prompt_type: MINIMAL, temperature: 0.7, budget: 0, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 1 |
| PLACED-07-7K | prompt_type: PERSONA, temperature: 0.7, budget: 7000, memory: continuous, replacement_round: 750, replacement_agents: 3,4, n_runs: 3 |
| RESET-07-7K | prompt_type: MINIMAL, temperature: 0.7, budget: 7000, memory: reset, memory_reset_rounds: 200,400,600,800, replacement_round: null, n_runs: 1 |
Factors: prompt_type (MINIMAL, PERSONA) × temperature (0.3, 0.7, 1, 1.4) × budget (3000, unlimited) × memory (continuous, reset)
Parameters
- n_agents
- 5
- n_rounds
- 1000
- model
- google/gemini-2.5-flash-lite
- scheduler
- round_robin