SAS Trustworthy AI Life Cycle
The SAS Trustworthy AI Life Cycle outlines steps for evaluating and deploying a more trustworthy AI system. The life cycle aims to make the U.S. National Institute of Standards and Technology's (NIST) recommendations, standards, and best practices for AI risk management easier to adopt and follow.
About
The life cycle helps organizational stakeholders specify individual roles and expectations, gather required documentation, and outline factors for consideration. As a result, teams can produce documentation to support the assertion that the organization has done its due diligence to provide evidence that the model is fair, and their processes do not cause harm.
The SAS Trustworthy AI Life Cycle is a stepwise process organized into sections. Many tasks in the life cycle reflect guidelines put forth by the NIST AI Risk Management Framework Playbook as it appeared in October 2023. Although the process is organized in a linear fashion, certain actions or outcomes might require returning to previous steps in the life cycle. Unsatisfactory outcomes might necessitate revisions in the data preparation and assessment step, the model training step, and beyond.
History
SAS originally created this resource to accompany the Trustworthy AI Life Cycle sample workflow that was orchestrated with SAS Workflow Manager on SAS Viya. The workflow provides the capability to easily make and manage responsible AI models from beginning to end. Although the steps in that sample workflow do not mirror the steps in this life cycle document in every respect, improvements to this document can result in requests for enhancements to the workflow. See the project README for additional details.
SAS seeks to expand responsible AI management practices, share expertise with other organizations, and spark innovation through community input. Anyone, anywhere, can use this document.
Contributing
Maintainers are accepting patches and contributions to this project. See CONTRIBUTING.md for details about submitting contributions.