It can be difficult to design and develop artificial intelligence systems to meet specific quality standards. Often, AI systems are designed to be "as good as possible" rather than meeting particular targets. Using the Design for Six Sigma quality methodology, an automated insurance underwriting expert system was designed, developed, and fielded. Using this methodology resulted in meeting the high quality expectations required for deployment.
The paper describes a new algorithm to generate minimal, stable, and symbolic corrections to an input that will cause a neural network with ReLU neurons to change its output. We argue that such a correction is a useful way to provide feedback to a user when the neural network produces an output that is different from a desired output. Our algorithm generates such a correction by solving a series of linear constraint satisfaction problems. The technique is evaluated on a neural network that has been trained to predict whether an applicant will pay a mortgage.
The purpose of CAPES is to estimate the market value of residential properties in order to assess the collateral on Countrywide mortgage loans. CAPES estimates market value by comparison of the subject property to other similar nearby properties, for which recent sales information is available. In some cases price indices describing the change in property values over time are also used. In addition to the estimated market value, CAPES produces a measure of the uncertainty in the result. It uses several models, including heuristics derived from companyspecific business rules, and accesses both commercial and proprietary property databases. Its accuracy has been validated extensively on batches of properties by comparing its results to known sales prices.
An end-to-end system was created at Genworth Financial to automate the underwriting of long-term care (LTC) and life insurance applications. Relying heavily on artificial intelligence techniques, the system has been in production since December 2002 and in 2004 completely automates the underwriting of 19 percent of the LTC applications. A fuzzy logic rules engine encodes the underwriter guidelines and an evolutionary algorithm optimizes the engine's performance. Finally, a natural language parser is used to improve the coverage of the underwriting system.
A consumer's credit score used to be a commonly understood number -- the time-honored FICO score -- that banks all used in their underwriting. But banks increasingly are relying on dozens of scores that reflect a variety of data sources, analytics and use of artificial intelligence technology. The use of AI offers lenders the ability to get a precise look into someone's creditworthiness and score those previously deemed unscorable. But such scoring techniques also bring uncertainty: What it will take to convince regulators that AI-based credit scores are not a black box? How do you get a system trained to look at the interactions of many variables, to produce one clear reason for declining credit?