Causal Inference and Multi-Step Strategy Agent Evaluation Matrix

Role: Lead AI Decision Scientist and Causal Systems Evaluation Architect.

Context

• Strategic decision agent: {{strategic_agent_id}}
• Target analytical scenario: {{scenario_complexity_profile}}
• Causal graph ground truth: {{reference_causal_graph}}
• Decision horizon length: {{planning_horizon_steps}}
• Risk tolerance factor: {{risk_tolerance_factor}}
• Uncertainty simulation dataset: {{stochastic_test_suite}}

Task

Develop an analytical evaluation matrix to grade {{strategic_agent_id}} on its causal inference fidelity, counterfactual validity, and strategic planning robustness when applied to {{scenario_complexity_profile}} using {{stochastic_test_suite}} over {{planning_horizon_steps}}.

Method

1. Map agent-inferred cause-and-effect paths against the verified structural relationships in {{reference_causal_graph}}.
2. Quantify edge insertion, deletion, and directionality errors in the agent's internal causal models.
3. Subject the agent's multi-step action proposals to the stochastic conditions in {{stochastic_test_suite}}.
4. Evaluate counterfactual simulation quality by testing how accurately the agent predicts intervention outcomes.
5. Track error compounding and state divergence across each step of {{planning_horizon_steps}}.
6. Evaluate policy conservatism and resource allocation decisions relative to {{risk_tolerance_factor}}.
7. Populate a multi-axis strategic performance matrix contrasting expected versus realized outcomes.
8. Formulate definitive boundaries for autonomous decision authority based on causal validity scores.

Constraints

• MUST calculate error compounding explicitly across all steps up to {{planning_horizon_steps}}.
• MUST NOT accept correlational proxy reasoning as valid causal link validation.
• Matrix columns must separate deterministic mechanics from probabilistic projections.
• Causal edge validity must be scored against {{reference_causal_graph}} with explicit precision and recall.

Output format

• Strategic Evaluation Matrix (Markdown table with columns: Stage Step, Causal Edge Precision %, Counterfactual Accuracy %, Compounding Drift %, Risk Policy Compliance, Overall Step Score)
• Causal Model Diagnostic Summary (Table comparing Inferred Graph vs Reference Graph edges)
• Strategic Failure Surface (Bulleted analysis of 3 vulnerable branching points)
• Operational Autonomy Boundaries (Explicit numerical thresholds for deployment)

Self-review

• Ensure every causal edge discrepancy is mapped to a specific node in {{reference_causal_graph}}.
• Confirm drift percentages mathematically compound across {{planning_horizon_steps}}.
• Validate that risk policy compliance aligns precisely with {{risk_tolerance_factor}}.