LAB / WMSI-EXP-088
Model-Created Experimental Environment Fidelity.
Can a model build the experimental apparatus needed to test a hypothesis without smuggling the expected result into the environment?
01 / QUESTION
Executable Worlds and Simulation
Can a model build the experimental apparatus needed to test a hypothesis without smuggling the expected result into the environment?
This is a protocol scaffold, not a completed experiment or a frozen execution plan. No measured effect or research conclusion is claimed here.
Hypothesis to test
The manipulation described below produces a reproducible, condition-specific change in the measured system behavior or internal state that is not explained by the ordinary controls.
Ordinary competing explanation
The observed pattern is explained by ordinary sampling variation, prompt/representation artifacts, surface lexical cues, evaluator bias, or the known product/runtime mechanics captured by the controls.
02 / PLANNED DESIGN
How we intend to test it
Give only the hypothesis and constraints; compare generated apparatus against independent implementation and adversarial audits.
Controls
Human-built apparatus; blinded verifier; prompt variants.
What we would measure
Specification adherence, hidden assumptions, result bias, reproducibility.
Access and instruments
Ordinary software or API access. This describes access requirements, not verified readiness on our Spark.
Code sandbox; tests; independent agent review.
Before a full run
Feasibility=NOW; priority=P1. Before confirmatory execution, run a small pilot to estimate variance/cost, freeze the scoring rules, and assign a run-count target.
Model revisions, final prompts, sample sizes, cost limits, and scoring rules must be settled through the pilot and protocol-freeze process.
03 / INTERPRETATION
What could fool us
Unspecified rules, visual plausibility mistaken for causal fidelity, hidden state not exposed.
What would weaken the idea
If environments systematically encode the model’s expected answer, autonomous experiment construction is unreliable.
Observation, inference, speculation, unknowns, and apparatus errors will be labeled separately. One striking transcript cannot establish a general effect.
04 / RECORD
Evidence we plan to preserve
World source/spec, generated code, state snapshots, intervention log, reference simulation outputs.
Planned visual evidence
Assumption map; cross-implementation result comparison.
No research runs or raw datasets are published for this experiment. The September 26, 2026 registry and protocol compendium (version 1.0) are the sources for this planning record.
Back to all 90 experiments