人工智能道德

Quick references

1. Overview

• Purpose: 

  • Provides a streamlined guide for organizations to adopt ethical AI practices in IT projects involving AI and big data analytics.

• Key Components:

  • Tailored AI Framework: Includes Ethical AI Principles, Governance Structure, Lifecycle, and Practice Guide.

  • AI Assessment: Tools to evaluate AI application impacts.

• Target Users: 

  • IT planners, system analysts, architects, data scientists, and executives.

2. Ethical AI Principles

• Performance Principles (foundational):

  • Transparency & Interpretability: 

    • Clear explanations of AI decisions.

  • Reliability, Robustness & Security: 

    • Consistent, secure, and error-resistant operations.

• General Principles (aligned with human rights and laws):

  • Fairness, Diversity & Inclusion, Human Oversight, Lawfulness, Data Privacy, Safety, Accountability, Beneficial AI, Cooperation & Openness, Sustainability.

3. AI Governance

• Three Lines of Defense:

  • Project Team: 

    • Develops and documents AI applications.

  • PSC/PAT: 

    • Reviews and approves projects (ensures quality).

  • IT Board/CIO + Ethical AI Committee:

    • Oversees high-risk projects and provides external advice.

4. AI Lifecycle

Six stages aligned with traditional software development but emphasizes iterative data refinement and feedback loops:

1. Project Strategy

2. Project Planning

3. Project Ecosystem

4. Project Development

5. System Deployment

6. Operation & Monitoring

5. AI Practice Guide Areas

Key practices across the lifecycle:

• Strategy/Planning: Align AI with organizational goals, comply with regulations.

• Development: Ensure data quality, mitigate bias, validate models.

• Deployment: Test integration, enable human oversight, monitor performance.

• Operation: Continuous review, compliance checks, stakeholder feedback.

6. AI Application Impact Assessment

• Process:

  • Use the Risk Gating Criteria to identify high-risk projects (requiring IT Board/CIO approval).

  • Complete the assessment at key lifecycle stages (e.g., planning, deployment).

• Components: Risk evaluation, ethical principle alignment, stakeholder impact analysis.

Key Actions for Adoption

• Apply all 12 Ethical AI Principles throughout projects.

• Align governance with the Three Lines of Defense.

• Follow the AI Practice Guide and complete Impact Assessments for each AI application.