Practice

Applying AI Ethics and Safety

1. Case Study Analysis

• Analyze real-world examples of AI misuse (e.g., biased hiring algorithms, facial recognition controversies, misinformation spread).

• Identify ethical concerns and propose ways to mitigate them.

• Example cases:

  • COMPAS Algorithm Bias in Criminal Justice

  • Amazon’s AI Recruiting Bias Issue

2. Bias Detection in AI Models

• Use an AI dataset and check for biases in training data.

• Apply fairness metrics using Python libraries (Fairlearn, AIF360).

• Compare model performance on different demographic groups.

3. AI Privacy and Security Assessment

• Conduct a privacy audit of an AI system:

  • What data is collected?

  • How is it stored and shared?

  • Are there compliance risks (GDPR, CCPA)?

• Simulate a data breach scenario and suggest mitigation steps.

4. Ethical AI Decision-Making Scenarios

• Provide hypothetical dilemmas where users must decide on AI-related ethical issues.

  • Should an AI system prioritize fairness over accuracy?

  • If an AI chatbot spreads false information, who is responsible?

  • Should an AI tool explain its decisions even if it makes it less effective?

5. AI Governance and Policy Writing

• Draft an AI Ethics Policy for a company adopting AI solutions.

• Include guidelines on bias reduction, transparency, and risk assessment.

6. Hands-On AI Explainability Task

• Take an AI model and use explainability tools (SHAP, LIME) to analyze why it makes certain decisions.

• Visualize decision-making patterns and discuss their ethical implications.