2022-10-20-白宫-美国白宫_AI权利法案蓝图_让自动化系统为美国人民服务（英）-2022.10_73页_730kb

BLUEPRINT FOR AN AI BILL OF RIGHTS Summary

The White House Office of Science and Technology Policy (OSTP) released the "Blueprint for an AI Bill of Rights" to guide the design, use, and deployment of automated systems in the United States, ensuring they align with democratic values and protect civil rights, civil liberties, and privacy. The blueprint outlines five key principles:

1. Safe and Effective Systems

• Automated systems must be tested, monitored, and designed to be safe and effective.
• Concerns and risks should be identified early through diverse community consultation.
• Data must be relevant and non-discriminatory, with protections against misuse or harm from data reuse.

2. Algorithmic Discrimination Protections

• Systems must avoid unfair discrimination based on race, gender, disability, or other protected classifications.
• Proactive equity assessments, representative data, and ongoing disparity testing are required to ensure fairness.

3. Data Privacy

• Built-in privacy protections must limit data collection to what’s strictly necessary.
• Individuals must have control over their data, including mechanisms for opt-out, deletion, and transparency.

4. Notice and Explanation

• Automated systems must provide clear explanations of how decisions are made, especially those affecting rights or opportunities.
• Ongoing transparency ensures users understand system functions and outcomes.

5. Human Alternatives, Consideration, and Fallback

• Users must be able to opt for human oversight when appropriate, especially in critical domains like justice, employment, or healthcare.
• Fallback processes should address failures or errors involving automated systems.

Key Features:

• Applied Scope: Covers automation beyond AI, including facial recognition, pre-trial risk assessments, and predictive algorithms.
• Non-Binding: Not law, but a framework guiding policymakers, industry, and developers.
• Informed Process: Development involved public engagement, addressing harms like biased algorithms and surveillance overreach.

The blueprint encourages technical and policy measures, backed by tools like companion guides, to institutionalize equity and accountability as U.S. societies adopt AI tools.