Recipe catalog / loss-cause-kind

Identify the cause of loss in a claim narrative

What cause of loss does narrative describe: weather, fire, water, theft, collision, wear and tear, vandalism, or something else?

A claims intake or routing system needs to sort a free-form loss description into a cause-of-loss category so it can pick the right adjuster queue, forms, or follow-up questions.

Explore this recipe interactively · Source and implementation guide

Use loss-cause-kind in TypeScript

Install with npm install jev-recipes. Requires Node.js 22.9 or newer and ES modules. Set TYPESAFE_API_KEY in your server environment for live calls, which send input to TypeSafe and use API quota. See the installation guide.

import { lossCauseKind } from 'jev-recipes/loss-cause-kind';

const result = await lossCauseKind({
  "narrative": "We returned from a long weekend on Monday evening and found the back door forced open, with the frame splintered around the deadbolt. Two laptops, my wife's jewelry box, a camera bag, and the cash we kept in the desk drawer were gone. Drawers were pulled out in every bedroom. We called the police that night and have a report number; the neighbor's doorbell camera shows two people in the yard at around 3 am on Sunday.",
  "minConfidence": 0.8
});
console.log(result);

Input contract

{
  "input": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "narrative": {
        "type": "string"
      },
      "minConfidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      }
    },
    "required": [
      "narrative"
    ]
  },
  "result": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "properties": {
      "model": {
        "type": "string"
      },
      "usage": {
        "type": "object",
        "properties": {
          "input_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          },
          "output_tokens": {
            "type": "integer",
            "minimum": 0,
            "maximum": 9007199254740991
          }
        },
        "required": [
          "input_tokens",
          "output_tokens"
        ],
        "additionalProperties": false
      },
      "status": {
        "type": "string",
        "enum": [
          "ready",
          "review"
        ]
      },
      "verdict": {
        "type": "string",
        "enum": [
          "weather",
          "fire",
          "water",
          "theft",
          "collision",
          "wear_and_tear",
          "vandalism",
          "other",
          "unclear"
        ]
      },
      "confidence": {
        "type": "number",
        "minimum": 0,
        "maximum": 1
      },
      "probabilities": {
        "type": "object",
        "propertyNames": {
          "type": "string",
          "enum": [
            "weather",
            "fire",
            "water",
            "theft",
            "collision",
            "wear_and_tear",
            "vandalism",
            "other",
            "unclear"
          ]
        },
        "additionalProperties": {
          "type": "number",
          "minimum": 0,
          "maximum": 1
        },
        "required": [
          "weather",
          "fire",
          "water",
          "theft",
          "collision",
          "wear_and_tear",
          "vandalism",
          "other",
          "unclear"
        ]
      }
    },
    "required": [
      "model",
      "usage",
      "status",
      "verdict",
      "confidence",
      "probabilities"
    ],
    "additionalProperties": false
  }
}

Saved example result

This hand-authored response demonstrates the contract. It is not a model accuracy measurement. Run it without an API key: npx jev-recipes demo loss-cause-kind.

{
  "model": "demo-fixture",
  "usage": {
    "input_tokens": 0,
    "output_tokens": 0
  },
  "status": "ready",
  "verdict": "theft",
  "confidence": 0.91,
  "probabilities": {
    "weather": 0.01,
    "fire": 0,
    "water": 0.01,
    "theft": 0.91,
    "collision": 0,
    "wear_and_tear": 0,
    "vandalism": 0.04,
    "other": 0.01,
    "unclear": 0.02
  }
}

Evaluation evidence

Use the evaluation guide to measure this decision on your own labeled cases.

Limitations

• Names the cause the narrative describes, not the cause an investigation would establish, and makes no coverage or exclusion determination.
• Returns the primary cause only. Chains such as a storm that causes a flood resolve to the initiating cause the narrative emphasizes, or to unclear when the narrative gives no dominant cause.

Related recipes

• failure-kind: Use failure-kind to classify why a technical operation failed rather than what caused an insured loss.
• shipment-issue-kind: Use shipment-issue-kind for what went wrong with a delivery; this recipe classifies the peril behind a property or vehicle loss.