AWS Certified AI PractitionerAI/ML and Generative AI FundamentalsHard
A healthcare provider is using a machine learning model to predict the likelihood of a patient developing a certain disease. They discover that the model consistently underpredicts the risk for a specific demographic group, leading to missed early interventions. This situation is an example of what ethical concern in AI?
- ALack of Interpretability
- BAlgorithmic Bias
- CData Leakage
- DOverfitting
Show answer & explanationAnswer & explanation
Correct answer: B. Algorithmic Bias
Algorithmic bias occurs when an AI system produces unfair outcomes for certain groups. In this case, the model 'consistently underpredicts the risk for a specific demographic group,' directly indicating that the algorithm is biased against that group, potentially due to biases in the training data or the model's design, leading to inequitable healthcare outcomes.
Why the other options are wrong
- A. Lack of interpretability means it's hard to understand how the model makes decisions, but the problem here is *what* the decision is (biased), not just *how* it's made.
- C. Data leakage is when information from outside the training data is inadvertently used to create the model, which doesn't directly describe unfair outcomes for a specific group.
- D. Overfitting refers to a model performing well on training data but poorly on unseen data generally, not specifically exhibiting unfairness towards a demographic group.
Algorithmic Bias
Systematic and repeatable errors in a computer system that create unfair outcomes, such as favoring one arbitrary group over another.
- Can stem from biased training data, flawed algorithm design, or skewed human feedback.
- Leads to discriminatory or unfair results for certain demographic groups.
- Manifests in various forms: underprediction, overprediction, misclassification.
Memory trick: AI's 'ethics' ensure 'fairness' and 'transparency' in its 'decisions'.