Computable Contracting Standards
reengineering of insurance contracts: InsurML and InsurLE.
InsurML: A standard to support the structured authoring of insurance contracts.
Axiome has developed an information model to support the modular representation of contracts. Once in modular form, a schema-based approach can be used to ‘semantically enrich’ each of the modular, contract components (e.g. through the addition of metadata). This enables a much higher level of control to be exercised over the creation and rendering of contracts. For example, it brings the notion of ‘mass customisation’ of contract wordings within reach, as well as the selection of individual contract components to support claims handling or customer queries. Semantic-web-based approaches have been successfully used for structured authoring in the number of domain-specific applications such as finance, medicine, publishing and aviation. The commercial insurance industry should be able to benefit by learning from implementations in other industry sectors.
InsurLE: A standard to express the rules embedded in an insurance contract explicitly, which can then be used to support logical reasoning.
An insurance contract is in essence a set of rules based on the promise of indemnifying the policyholder in the event of loss. This is written in natural language form in the contract wording as something like: ‘we will cover costs in respect of a loss provided that specific conditions are met’. Today, these rules are often hard-coded into systems e.g. to support claims handling. Crucially, the rules in the contract wording and the rules expressed by the code in business systems are fundamentally divorced from one another. This leads to problems on whether the code is an accurate, real-time representation of the contract, particularly when wordings are subject to a constant stream of amendments. With computable contracts, the rules may be expressed in a controlled natural language: high-level code that actually looks like natural language, but is actually a codified version of the rules. This approach helps to establish a much closer bond between the contract wordings and the actionable logic (in the form of a computer program). Under the banner of ‘neurosymbolic AI’ (or the ‘third wave in AI’), the use of these deterministic, rule-based approaches (based upon transparent reasoning), together with the powerful (but fundamentally probabilistic) approaches of large language models, is securing significant levels of interest.