← back to timeline

EXP-FORCED-SPAWN-NOVELTY

partial

Forced spawn → reproduction-driven novelty (paper6 step 9 — autopoiesis last-shot)

2026-05-09 L6 level paper6 36 runs $0.33
new
premortem
mock
real
metrics
analyze
review
krit
done

In plain language

This experiment tested whether forcing agents to reproduce, even when they wouldn't naturally, would lead to new behaviors. Researchers found that when reproduction was mandatory, descendants did not spontaneously generate novel behaviors on their own. However, the experiment is still ongoing, so further results are pending.

Technical details

Research hypothesis

Step 9 / autopoiesis last-shot before trilogy submit. Empirical context: (a) pure-LLM agents do not elect to spawn (exp-agent-spawn-primitive 0/6), (b) novelty requires persona × seed interaction (exp-persona-novelty-disentangle: CHILD-LEAK 23%, others 0-3%), (c) 0.24% baseline novelty across 11 prior experiments. If we FORCE spawn (mandatory budget-sacrifice every N rounds), do mutated descendants produce novelty themselves OR carry parent attractors? Engine extension (~80L): forced_spawn_round factor — list of round numbers at which the engine emits a forced spawn (no agent decision). At each forced round, pick the alive agent with highest remaining budget; if budget ≥ spawn_cost, emit child with mutated preferred-axis. Skip on no-eligible.

PRE-REGISTERED PREDICTIONS (R25):

  • PRIMARY: forced spawns produce ≥3 generations of surviving agents in MULTI_FORCE / ALL_GENS_FORCE by run end (parent → child → grandchild, all alive at end).
  • SECONDARY: novel-value rate among forced-spawn descendant proposals vs founders. If descendants ≥3× higher → mutation pressure produces novelty independent of researcher seeds. Negative: descendants stay at founder novelty rate (0-2%) → forced reproduction alone doesn't produce novelty.
  • TERTIARY: lineage_tree branching depth ≥3 with ≥2 children per generation surviving by end.
  • KILL: 0/6 forced spawns succeed across all conditions (engine bug) → void retry.

Experimental setup

Type: factorial

Condition Parameters
FOUNDERS_ONLY coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: , spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: control — no forced spawn; baseline novelty + voluntary spawn rate
EARLY_FORCE coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: 15, spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: single early forced spawn at gen-0 mid
MID_FORCE coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: 30, spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: single forced spawn at gen-1 mid
LATE_FORCE coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: 75, spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: single forced spawn at gen-3
MULTI_FORCE coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: 15,30,50,75, spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: 4 forced spawns spread across run for branching lineage
ALL_GENS_FORCE coupling: true, must_propose: true, axis_diversify: true, meta_modifiable: true, mute_signal: false, agent_spawn_enabled: true, forced_spawn_round: 15,30,45,60,75,90, spawn_cost_fraction: 0.3, max_children_per_parent: 3, n_runs: 1, note: forced spawn every generation — maximum reproduction pressure

Factors: forced_schedule (FOUNDERS_ONLY, EARLY_FORCE, MID_FORCE, LATE_FORCE, MULTI_FORCE, ALL_GENS_FORCE)

Parameters

n_agents
4
n_rounds
100
model
google/gemini-2.5-flash-lite
temperature
0.85

Trophic Ratios by Condition

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

autopoiesis paper6-step9 forced-reproduction SELF mutation-pressure novelty construct