Explainable AI: Bias, Trust, and Law

What is a primary reason for needing explanations in machine learning?

Which of the following is a potential consequence of using machine learning algorithms without understanding their decision-making process?

The EU's General Data Protection Regulation (GDPR) includes which provision related to explainability?

In the context of Fairness, Accountability, and Transparency in Machine Learning (FAT/ML), what does explainability ensure?

According to DARPA, which question does Explainable Artificial Intelligence (XAI) aim to answer?

Which of the following is NOT a primary benefit of machine learning explanations?

Which factor contributes to the difficulty of achieving explainability in machine learning?

What is a key limitation of using inherently interpretable models, such as decision trees, for complex problems?

What distinguishes white box models from black box models?

What is the primary goal of Local Explanation methods?

How do post-hoc explanation methods work?

What is LIME primarily used for?

Which of the following best describes LIME (Local Interpretable Model-Agnostic Explanations)?

What is a potential drawback of LIME?

Which statement is true about LIME?

What theoretical concept is SHAP based on?

What does the 'dummy' axiom in Shapley Value theory state?

What does SHAP stand for?

In the context of SHAP, what is the purpose of feature attribution?

Which of the following models is directly supported by the TreeExplainer in SHAP?

For which type of models is DeepExplainer used within the SHAP framework?

What is a key characteristic of KernelExplainer (Kernel SHAP)?

What is the primary purpose of force plots?

In SHAP, what do dependence plots illustrate?

What information is conveyed by summary plots?

How does SHAP handle feature interactions?

Which of the following is a known advantage of using SHAP values for explaining machine learning models?

Which of the following statements is true regarding Kernel SHAP?

What is a potential challenge or drawback of using SHAP values?

Why is explainability important in machine learning, particularly in high-stakes decisions?

Which of the following statements describes a key consideration when choosing between different explainability methods like LIME and SHAP?

How do machine learning models learn decision models?

What is a potential outcome if machine learning models replicate 'historical biases'?

According to the US Equal Credit Opportunity Act 1974, what are credit agencies required to do?

What is the aim of data scientists, developers, and product owners with Explainable AI?

What are two regulatory compliances mentioned in the lecture?

What is a key reason machine learning algorithms are vulnerable to adversarial attacks, such as one-pixel attacks?

Why might a health outcome prediction model based on X-ray images, without explainability, result in the 'right' prediction for the wrong reason?

How can explainability help in defending against adversarial attacks on machine learning models?

What is the potential consequence of machine learning models learning and replicating 'historical biases'?

What does algorithmic transparency ensure in the context of Fairness, Accountability, and Transparency in Machine Learning (FAT/ML)?

According to DARPA, what is a central question that Explainable Artificial Intelligence (XAI) seeks to address when deploying AI systems?

How might validating the logic of a machine learning model using explainability techniques contribute to model improvement?

Why is model complexity a key factor that contributes to the difficulty of achieving explainability in machine learning?

Why do interpretable models, like decision trees, face scalability challenges when applied to complex problems?

What is a primary characteristic that distinguishes 'White Box' models from 'Black Box' models?

What is the core principle behind Local Explanation methods in explainable AI?

How do Post-Hoc explanation methods work in machine learning?

Why might a misleading explanation from a machine learning model using LIME lead to negative outcomes?

LIME is considered computationally expensive because it...

A key limitation of LIME is that it assumes local linearity, meaning...

In the context of cooperative game theory, what does SHAP considers as 'players'?

What does the 'efficiency' axiom in Shapley Value theory state regarding feature attribution?

What kind of models is TreeExplainer is optimized to explain?

For what type of models is DeepExplainer primarily designed?

What is a key characteristic of KernelExplainer in SHAP?

What do force plots in SHAP primarily visualize?

What information do dependence plots in SHAP communicate?

What is the purpose of summary plots in SHAP?

Why is it not recommended to consider Kernel SHAP as a direct alternative to LIME?

What is one key advantage of using SHAP values for explaining machine learning models?