Our People and Culture 98 2021 Sustainability Report Responsible Business Practices A Modern, High-Performing Health System Introduction Performance Data Environmental Health Our principles The health care industry as a whole is in the early stages of implementing AI/ML. As AI/ML solutions continue to evolve, it is important to ensure they are used responsibly by continuously assessing and improving our governance processes. We adopted the following principles as guideposts to develop, deploy and monitor our AI/ML solutions. Mission-driven Foster development and deployment of AI/ML and governance processes consistent with our mission. Trust Employ methods to test and monitor AI/ML integrity and reliability. Fairness Create procedures to assess AI/ML performance for possible bias. Accountability Establish measures and be prepared to act swiftly to address and remediate misuses or adverse outcomes. Transparency Enable reviews of data and AI/ML outputs. Privacy Safeguard data privacy in the design, deployment and use of AI/ML. Governance We established a governance structure consisting of leaders across UnitedHealth Group, UnitedHealthcare and Optum to provide oversight on strategy development and guidance for investments and capability development of AI/ML. This governance established foundational principles for the responsible use of AI/ML and a Responsible Use of ML program. The program is designed to provide a holistic and structured approach to using AI/ML responsibly. In 2022, we will integrate the Machine Learning Review Board as a key component of the Responsible Use of ML program. The board will proactively review AI/ML models for bias and similar issues and advise analytics teams. The program will provide guardrails in the application of innovative solutions that advance the health care system. As we move forward, we will supplement our foundational principles with a guide for all analytics teams to support consistent practices to responsibly develop and use AI/ML. The Responsible Use of ML program will make this guide available to all employees, along with training. We will develop a technology framework that standardizes and automates AI/ML quality checks for use across our enterprise. Finally, we are engaging with industry and academic partners to review our best practices and to co-research and publish leadership for the benefit of the health industry. We recognize that the use of AI/ML can have unintended consequences, including consequences that can arise from bias in the health care system, data or algorithms. We take steps throughout the design, development, deployment and monitoring of AI/ML to help mitigate bias. These steps include forming a team with a range of expertise to help identify potential issues, assessing the data and model for potential sources of bias and developing cross-discipline mitigation approaches to mitigate bias. We are developing processes for situations where bias or risks to health equity are found that may include suspending or delaying release of AI/ML until the risks are addressed through a mitigation plan.