The purpose of this course is to equip learners with the knowledge and practical skills to design, implement, and evaluate artificial intelligence systems that are not only effective but also transparent, accountable, and ethically responsible. By engaging with concepts of explainability, interpretability, and fairness, learners will develop the capacity to critically assess AI models, apply explainable AI techniques, and communicate insights in a manner that fosters trust among diverse stakeholders. The course further emphasizes responsible AI practices by addressing issues of bias, accountability, legislative frameworks, and sustainability, ensuring that graduates can contribute to the development of AI systems that are both technically sound and socially aligned.

CLO1. Discuss the objectives of explainability in artificial intelligence solutions

CLO2. Implement explainability in AI techniques in a selected programming language

CLO3. Evaluate the explainability of machine learning models

CLO4. ACommunicate explainable AI results to stakeholders

• Introduction to Explainable AI:
  AI Model Complexity; Explainability; Interpretability; Transparency; Accountability; Bias; Fairness; Legislative Framework; Ethical Considerations; Green Computing; Data Protection
  
• Forms of Explanation:
  Analytical Explanation; Visual Explanation; Rule-based Explanation; Textual Explanation
  
• Model-Specific Explainability:
  Rule-based learner; Decision Tree; Feature Relevance
  
• Model Agnostic Explainability:
  Explanation by Simplification; Rule-based learner; Decision Tree; Feature Relevance; Influence Functions; Sensitivity; Local Interpretable Model Agnostic Explanations (LIME); Shapley Values (SHAP); Visual Explanations; Dependence Plots; Sensitivity; Local Explanation; Rule-based learner; Linear Approximation; Counterfactuals
  
• Explainable AI Implementation:
  Explainable AI Libraries: SHAP; SHAP Values; SHAP Value Plots and Interpretations; LIME Module; LIME Value Plotting and Explanation; Decision Tree Plotting; Shapash (Metrics, Plots, Explanation)
  
• Explainable AI Evaluation:
  Confidence; Fidelity; Stability; Comprehensibility