人工智慧治理與可信任設計
Governance and Trustworthy Artificial Intelligence
| 節 | 週二 |
|---|---|
2 09:00–09:50 | 人工智慧治理與可信任設計 ED201 3 節連堂 |
3 10:10–11:00 | |
4 11:10–12:00 |
* 根據陽明交大上課時間表所列
本課程介紹人工智慧系統的治理架構與可信任設計原則，強調資料、模型與系統決策在不同階段對公平性、透明性、隱私與安全所造成的影響。課程內容包含人工智慧治理 (AI Governance)、責任導向設計、公平與偏誤分析、模型透明與可解釋性、隱私保護與系統安全等主題。學生將學習如何評估與設計具可信任特性的人工智慧系統，並透過實作深入理解技術選擇如何影響使用者信任與社會影響。
Introduction to Artificial Intelligence/Deep Learning/Machine Learning, Probability
Assignment *3 (45%) Individual-based Project *1 (20%) Team-based Project *1 (25%) Class Participation (10%)
| 週次 | 主題 |
|---|---|
| 第 1 週 | Introduction to Governance and Trustworthy AI |
| 第 2 週 | Data Ethics (I) - Deployed AI technologies |
| 第 3 週 | Data Ethics (II) - Ethical and Social Issues of Data [HW1: Data Ethics Case Analysis] |
| 第 4 週 | Data Ethics (III) - Data Governance & Accountability |
| 第 5 週 | Bias and Fairness (I): Definitions and Perspectives |
| 第 6 週 | Bias and Fairness (II): Measuring and Analyzing Bias [HW2: Fairness Evaluation Exercise] |
| 第 7 週 | Bias and Fairness (III): Mitigation and Design Choices |
| 第 8 週 | Project Session |
| 第 9 週 | Privacy (I): Privacy Risks and Trade-offs |
| 第 10 週 | Privacy (II): Privacy-Preserving AI [HW3: Privacy Risk Reflection] |
| 第 11 週 | Project Session [Individual-Project Proposal Due] |
| 第 12 週 | Security, Privacy & Safety of AI |
| 第 13 週 | Designing Trustworthy AI Systems [Team-Project Proposal Due] |
| 第 14 週 | Project Session [Individual-based Project Presentation] |
| 第 15 週 | |
| 第 16 週 | Project Session [Team-based Project Presentation] |
| 第 17 週 |
Breyer, Sabrina, and Christian Herzog. "Integrating ethical considerations into innovation design." Novel innovation design for the future of health: Entrepreneurial concepts for patient empowerment and health democratization. Cham: Springer International Publishing, 2022. 253-282. https://link.springer.com/chapter/10.1007/978-3-031-08191-0_23#Sec16 Vashney, Kush R. Trustworthy machine learning. Independently published, 2022. http://www.trustworthymachinelearning.com/trustworthymachinelearning.pdf