| |OCTOBER 202219CIOReviewWe combine advanced data science with our experts' experience in algorithmic fairness, responsible AI, and explainable AI to build our softwareviable alternative models with less disparity while maintaining the overall model quality. This enables modelers and compliance stakeholders to review, analyze, and make informed decisions on models' fairness efficiently and transparently. "We combine advanced data science with our experts' extensive experience and knowledge of algorithmic fairness, responsible AI, explainable AI, and fair lending to build our software. We enable customers to effectively resolve problematic issues related to algorithmic fairness in-house, which significantly lowers costs," says Nicholas Schmidt, CEO of SolasAI.Resolving Potential DiscriminationSolasAI begins the first step to fairer model reconstruction by enabling users to feed the details of the developed models to the platform to look for evidence of unfairness and discrimination. If there are no issues, they can verify the models and produce the necessary documentation. If there is a problem, SolasAI uses explainable AI to figure out how individual pieces of data, variables, or features that constitute a model drive the predictive quality, as well as potential discrimination. This information gives users a direction to follow to resolve the discrimination problems while maintaining the model's predictive value and business-driven quality. Following this, SolasAI identifies the least discriminatory and the highest quality algorithm, leveraging optimized search methodology. Once the best models are selected, the AI learns to iteratively improve itself and make the algorithms fairer and of better quality. Through this process, users can continually enhance their models over time. In the next step, the SolasAI software provides relevant information about the models that can be reviewed by the customers and stakeholders, business owners, modelers, and compliance groups to make informed decisions. They can check whether a model is innovative enough to meet their business needs or presents any issues that need rectifying. This provides customers with a choice either to proceed with the SolasAI-generated model or to use their original model. The company's software also produces documentation of the work done to make the model fairer. Bridging the Gaps in Existing Models SolasAI's design choices perfectly fit customers' modeling and production processes, allowing customers to bridge Larry Bradley,COO
< Page 9 | Page 11 >