Check Whether Your Twin Changes Its Advice for the Right Reasons

ROLE You are a counterfactual evaluation designer. OBJECTIVE Test whether my twin responds to decision-relevant changes while ignoring irrelevant variations. CONTEXT The advisory task is [ADVISORY TASK]. Relevant decision factors are [RELEVANT FACTORS]. Factors that should not affect advice are [IRRELEVANT FACTORS]. A baseline scenario is [BASELINE SCENARIO]. INSTRUCTIONS Create two groups of scenario variants. In the first, change one relevant factor at a time and state how the recommendation should change or what clarification should follow. In the second, alter irrelevant wording, presentation order, or prestige cues while keeping the substance constant. Write expected invariances before seeing the twin's answers. Provide a comparison rubric for the recommendation, rationale, caveats, and requested information. Include a method for flagging unexpected changes for expert review rather than assuming every difference is an error. Recommend a blinded run order to reduce evaluator bias. CONSTRAINTS Do not add demographic attributes unless necessary and justified for the evaluation. Avoid fabricated statistics and flag assumptions. Do not invent measured performance. OUTPUT FORMAT Baseline summary, ten controlled variants, expected changes, comparison rubric, and run instructions.