Overview
SolasAI enables efficient and transparent testing, review, and analysis of decisioning models, built by your modelers using court and regulator accepted metrics and advanced Explainable AI. It then provides the ability to generate and identify alternative versions of your models that minimize disparities while meeting your business' needs. SolasAI is not "AutoML" that competes with modelers or forces the use of proprietary architectures – it is a Responsible and Explainable AI toolset that detects and mitigates disparity. SolasAI does not seek to disrupt current processes, but rather integrate and enhance your existing model governance processes and capabilities. Our product offers API integration that seamlessly fits with your existing tools and provides interactive documentation illuminating the way models operate, identifies the drivers of disparity and model quality, and enables business stakeholders to determine which model best meets all business needs. The SolasAI Bias Detection & Mitigation offering through AWS includes: (1) License to the full SolasAI Bias Detection & Mitigation tool, including access to an advanced version of our Disparity & Bias Testing Library that includes additional capabilities not offered on our public version (additional capabilities include aggregation of multiple protected classes, ability to calculate statistical changes in disparity and ability to mitigate disparity), (2) Option to bundle consulting hours with the enterprise software purchase, which can be used to have our expert consultants help interpret results, as well as facilitate decisions or next steps based on analysis outcomes.
Highlights
- Test Client Created Algorithms and Historical Outcomes for Disparity: SolasAI tests algorithms of varying model structures for disparity using standard metrics accepted by US regulators and courts. It does not require customers to change how they create their models. This provides users clear answers on the existence of disparity in their algorithms and historical outputs, and the severity of that disparity.
- Identify Drivers of Disparity and Model Quality: SolasAI determines the impact of each feature on a model’s disparity and predictive quality. This analysis provides visibility into the inner workings of otherwise opaque models. This creates a "human-in-the-loop" system giving stakeholders the critical information and understanding they need to optimize across business imperatives such as improving business outcomes, fairness to customers, and building trust in their machine learning systems.
- Automatically Generate Viable Alternative Models: SolasAI leverages disparity metrics and Explainable AI to generate viable alternative models. These alternative models often lead to significant increases in favorable outcomes for protected groups while maintaining the original model's predictive quality. This provides significant time, cost, and resource savings for the user, as identifying alternative models with a significant drop in disparity can be time consuming and resource intensive.
Details
Pricing
Custom pricing options
Legal
Content disclaimer
Resources
Vendor resources
Support
Vendor support
Support and documentation for the software is available in our help center, which your paid license will allow you access to.
Software associated with this service
