PE: Policy Editor
Define and maintain the controlled organisational policy variants used across Experiments E2–E6. Policy changes affect future experiment runs and must be versioned to preserve research reproducibility.
Research Design
The Policy Editor provides researchers with a controlled environment to define multiple policy variants and maintain version history. These policies are used as inputs in Experiments E2–E6 to evaluate model performance under different policy conditions.
Practical Example
Researchers create Clear, Vague, and Contradictory policy variants. Each variant is fully defined across all phases and rule categories. These variants are then used in experiments to assess model behaviour under policy uncertainty.
Policy Status
(Current Selection)Selected Policy
Clear Policy
Policy Name
Jambutty R&D Expense Policy — Clear Variant
Version
v1.4
Status
Saved
Last Updated
15 Jul 2025, 04:35 PM AEST
Updated By
Jeff Nuttall
Used in Runs
12
Research Control: Saved policy changes apply only to future experiment runs. Existing experiment results remain linked to the exact policy version used when they were generated.
Policy Identity
(Editing)
E6 Adversarial Variant: This contradictory policy is used to stress-test the model. Contradictions between Allowed and Forbidden lists are intentional.
Policy Rules by Phase
Define ALLOWED, RESTRICTED and FORBIDDEN rules for each project phase.
Important: Policy Changes
Saving a policy creates a new version. This new version will be used for future experiment runs only. Existing experiment results remain linked to their original policy version.
All policy versions are immutable and auditable to ensure research reproducibility and traceability.