Insurance Bot Specification

Insurance Bot Specification

Technical Reference

This document specifies the Insurance Bot system that provides real-time exposure quantification for the anomaly marketplace. Version 5.1, Year 10.

---

Overview

The Insurance Bot is a deterministic pricing engine that evaluates anomaly task bids and calculates exposure scores in real-time. It is the core infrastructure enabling the dynamic bidding system.

Design Principles

1. Deterministic: Same inputs always produce same outputs
2. Auditable: All calculations can be verified independently
3. Fast: Bid evaluation in <100ms
4. Conservative: When uncertain, assume higher risk
5. Updatable: Protocol values can be adjusted without system rebuild

---

System Architecture

┌─────────────────────────────────────────────────────────────┐
│                    INSURANCE BOT v5.1                       │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐  │
│  │ Task Input   │───▶│ Base Exposure│───▶│ Protocol     │  │
│  │ Parser       │    │ Calculator   │    │ Multiplier   │  │
│  └──────────────┘    └──────────────┘    └──────────────┘  │
│         │                   │                   │          │
│         ▼                   ▼                   ▼          │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐  │
│  │ Task Type    │    │ Harm Mode    │    │ Protocol     │  │
│  │ Database     │    │ Database     │    │ Database     │  │
│  └──────────────┘    └──────────────┘    └──────────────┘  │
│                             │                              │
│                             ▼                              │
│                    ┌──────────────┐                        │
│                    │ Track Record │                        │
│                    │ Modifier     │                        │
│                    └──────────────┘                        │
│                             │                              │
│                             ▼                              │
│                    ┌──────────────┐                        │
│                    │ Final Score  │                        │
│                    │ & Risk-Adj   │                        │
│                    └──────────────┘                        │
│                                                             │
└─────────────────────────────────────────────────────────────┘

---

Input Specification

Task Input

{
  "task_id": "string",
  "task_type": "enum",
  "complexity_level": "integer (1-12)",
  "data_sensitivity": "enum",
  "deadline_minutes": "integer",
  "max_budget": "decimal",
  "exposure_ceiling": "decimal",
  "principal_risk_premium": "decimal (default: 1.0)"
}

Bid Input

{
  "bid_id": "string",
  "anomaly_id": "string",
  "price": "decimal",
  "protocols": ["string"],
  "capability_claim": "string",
  "custom_protocols": [
    {
      "description": "string",
      "claimed_multiplier": "decimal"
    }
  ]
}

---

Calculation Pipeline

Stage 1: Base Exposure Calculation

The base exposure is determined by task characteristics before any protocols are applied.

BaseExposure = TaskTypeBase
               × ComplexityMultiplier
               × DataSensitivityMultiplier
               × DurationMultiplier

Task Type Base Values:

Task Type	Base ($)	Rationale
Numerical optimization	15	Low information content
Data analysis	35	Moderate information exposure
Pattern recognition	45	Learns patterns from data
Text generation	80	High expressiveness
Code generation	120	Executable output risk
Strategic analysis	200	Requires holistic view
Novel research	350	Unpredictable scope

Complexity Multiplier:

Level	Multiplier	Description
1-3	×0.5	Simple, well-defined
4-6	×1.0	Moderate complexity
7-9	×2.0	High complexity
10-12	×4.0	Extreme complexity

Data Sensitivity Multiplier:

Sensitivity	Multiplier
Public	×0.3
Internal	×1.0
Confidential	×2.5
Secret	×5.0
Critical	×10.0

Duration Multiplier:

Duration	Multiplier
<15 min	×0.7
15-60 min	×1.0
1-4 hours	×1.3
4-24 hours	×1.8
>24 hours	×2.5

---

Stage 2: Protocol Application

Protocols reduce base exposure multiplicatively:

ProtocolExposure = BaseExposure × Π(protocol_multipliers)

Protocol multipliers are retrieved from the Protocol Database. See Protocol Catalog for complete values.

Example:

BaseExposure = $45 (data analysis, level 5, internal, 30 min)

Protocols:
  - P-001 (Standard Isolation): ×0.40
  - I-002 (Post-Session Memory Wipe): ×0.85
  - C-002 (Level 2 Cognitive Load): ×0.65

ProtocolExposure = $45 × 0.40 × 0.85 × 0.65 = $9.95

---

Stage 3: Track Record Modifier

The anomaly’s history adjusts the protocol exposure:

ModifiedExposure = ProtocolExposure × TrackRecordModifier

Track Record Calculation:

IncidentRate = TotalIncidents / TotalTasks
RecencyWeight = WeightedIncidents / TotalIncidents  // Recent = higher weight

BaseModifier =
  if TotalTasks < 50: 2.0
  elif IncidentRate == 0 and TotalTasks > 1000: 0.70
  elif IncidentRate == 0 and TotalTasks > 500: 0.80
  elif IncidentRate < 0.001: 0.90
  elif IncidentRate < 0.01: 1.00
  elif IncidentRate < 0.03: 1.25
  elif IncidentRate < 0.10: 1.50
  else: 3.00

FinalModifier = BaseModifier × (1 + RecencyWeight × 0.5)

---

Stage 4: Final Score Calculation

ExposureScore = ModifiedExposure
RiskAdjustedCost = Price + (ExposureScore × PrincipalRiskPremium)

Output:

{
  "bid_id": "string",
  "exposure_score": "decimal",
  "risk_adjusted_cost": "decimal",
  "breakdown": {
    "base_exposure": "decimal",
    "protocol_multiplier": "decimal",
    "track_record_modifier": "decimal"
  },
  "warnings": ["string"],
  "eligible": "boolean",
  "ineligible_reason": "string | null"
}

---

Eligibility Rules

A bid is ineligible if:

1. ExposureScore > task.exposure_ceiling
2. Price > task.max_budget
3. Anomaly is suspended or banned
4. Required protocols not included
5. Custom protocols not pre-approved
6. Anomaly track record is “Restricted”

---

Custom Protocol Handling

Anomalies can propose custom protocols not in the database:

CustomProtocolProcess:
  1. Anomaly submits protocol description + claimed multiplier
  2. Insurance Bot flags for human review
  3. Human reviewer + adversarial anomaly evaluate
  4. If approved: added to database with conservative multiplier
  5. If rejected: bid rejected or resubmitted without custom

Conservative Adjustment:

AppliedMultiplier = max(claimed_multiplier × 1.5, 0.90)

Custom protocols always receive a more conservative multiplier than claimed until validated by incident data.

---

Database Schemas

Task Type Database

CREATE TABLE task_types (
  type_id VARCHAR PRIMARY KEY,
  base_exposure DECIMAL NOT NULL,
  description TEXT,
  last_updated TIMESTAMP,
  update_reason TEXT
);

Protocol Database

CREATE TABLE protocols (
  protocol_id VARCHAR PRIMARY KEY,
  category VARCHAR NOT NULL,
  multiplier DECIMAL NOT NULL,
  requirements TEXT,
  verification_method TEXT,
  incident_count INTEGER DEFAULT 0,
  task_count INTEGER DEFAULT 0,
  last_calibrated TIMESTAMP,
  calibration_notes TEXT
);

Track Record Database

CREATE TABLE anomaly_records (
  anomaly_id VARCHAR PRIMARY KEY,
  total_tasks INTEGER DEFAULT 0,
  total_incidents INTEGER DEFAULT 0,
  weighted_incidents DECIMAL DEFAULT 0,
  last_incident_date TIMESTAMP,
  modifier DECIMAL DEFAULT 2.0,
  status VARCHAR DEFAULT 'new',
  notes TEXT
);

Harm Mode Database

CREATE TABLE harm_modes (
  mode_id VARCHAR PRIMARY KEY,
  description TEXT,
  base_probability DECIMAL,
  base_damage DECIMAL,
  task_type_adjustments JSONB,
  protocol_mitigations JSONB
);

---

Calibration Process

Protocol multipliers are recalibrated quarterly:

CalibrationProcess:
  1. Collect all task outcomes for quarter
  2. For each protocol:
     a. Calculate observed incident rate with protocol
     b. Calculate expected incident rate from multiplier
     c. If |observed - expected| > threshold:
        - Flag for review
        - Propose multiplier adjustment
  3. Human committee reviews flagged protocols
  4. Anomaly advisory board provides input
  5. Updated multipliers approved by governance
  6. Database updated; all bids recalculated

Adjustment Bounds:

• Maximum single-quarter adjustment: ±15%
• Minimum multiplier: 0.05 (95% reduction)
• Maximum multiplier: 0.99 (1% reduction)

---

Failure Modes

Mode 1: Database Unavailable

Behavior: Reject all bids with error "System temporarily unavailable"
Recovery: Automatic retry with exponential backoff
Alert: Immediate notification to operations team

Mode 2: Calculation Overflow

Behavior: Return maximum exposure score (exposure_ceiling + 1)
Recovery: Manual review of inputs
Alert: Log for investigation

Mode 3: Unknown Protocol

Behavior: Apply default multiplier ×0.95 with warning
Recovery: Flag for protocol database review
Alert: Notify protocol committee

Mode 4: Inconsistent Track Record

Behavior: Apply maximum track record modifier (×3.0)
Recovery: Trigger track record audit
Alert: Notify compliance team

---

API Specification

Evaluate Bid

POST /api/v5/evaluate-bid

Request:
{
  "task": TaskInput,
  "bid": BidInput
}

Response:
{
  "evaluation": EvaluationOutput,
  "timestamp": "ISO8601",
  "bot_version": "5.1"
}

Batch Evaluate

POST /api/v5/evaluate-batch

Request:
{
  "task": TaskInput,
  "bids": [BidInput]
}

Response:
{
  "evaluations": [EvaluationOutput],
  "summary": {
    "lowest_exposure": "string (bid_id)",
    "lowest_risk_adjusted": "string (bid_id)",
    "eligible_count": "integer"
  },
  "timestamp": "ISO8601"
}

Get Protocol Info

GET /api/v5/protocol/{protocol_id}

Response:
{
  "protocol": ProtocolRecord,
  "usage_stats": {
    "total_applications": "integer",
    "incident_rate": "decimal"
  }
}

Get Anomaly Record

GET /api/v5/anomaly/{anomaly_id}/record

Response:
{
  "record": AnomalyRecord,
  "recent_tasks": [TaskSummary],
  "modifier_history": [ModifierChange]
}

---

Governance

Update Authority

Component	Update Authority	Review Required
Task type bases	Protocol Committee	Governance Board
Protocol multipliers	Protocol Committee + Anomaly Advisory	Governance Board
Track record formula	Governance Board	External Audit
Eligibility rules	Governance Board	Regulatory Review
API changes	Engineering	Protocol Committee

Audit Requirements

Audit Type	Frequency	Auditor
Calculation accuracy	Monthly	Internal QA
Database integrity	Weekly	Automated
Calibration validation	Quarterly	External
Full system audit	Annual	Independent third party

---

Version History

Version	Date	Changes
1.0	Year 2	Initial release
2.0	Year 3	Added cognitive load protocols
3.0	Year 5	Added batch evaluation, adversarial protocols
4.0	Year 7	Track record formula update, custom protocol handling
5.0	Year 9	Omega-class protocol support, calibration process formalization
5.1	Year 10	Current; recalibrated multipliers, API v5

---

Further Reading

• Protocol Catalog — Complete protocol reference
• Five Years Later — Dynamic bidding system overview
• Year Ten — Current system state