1. privacy and security - data minimization, accuracy 2. fairness and inclusion - misinformation in using ai in healthcare, dissecting racial bias in alg 3. robustness and safety - does the system do what it says it does - does the system cause harm, directly or indirectly 4. transparency and control - a.k.a. explainability, interpretability - for safety, customization, regulation compliance 5. accountability and governance - who is responsible for decision, ensuring compliance for decision, uber reaches sett ## stakeholder impact analysis | stakeholers | potential system benefits | potential system harms | | ----------- | ------------------------- | ---------------------- | | 1. | | | | 2. | | | | 3. | | | ![[Pasted image 20230809154448.png]] ## Mitigating risks for gAI ## lab + qna lab for zero-shot, training prompting as introduced in [here](https://www.aicareerboost.com/products/complete-ai-product-leader-blueprint/categories/2153148756/posts/2170131019) ![[Pasted image 20230809192206.png]] ![[Pasted image 20230809192259.png]] - accuracy is not deterministic