AWS Certified AI PractitionerResponsible AIHard
A credit scoring AI model is found to consistently give lower scores to individuals residing in certain postal codes, even when controlling for other financial indicators. Upon investigation, it's discovered that these postal codes historically correspond to lower-income areas, and the training data implicitly learned to associate location with creditworthiness, regardless of individual applicant merits. This leads to unfair outcomes for a specific demographic group. Which type of Responsible AI issue is this model exhibiting?
- AAI privacy breach
- BLack of AI interpretability
- CAI robustness failure
- DAlgorithmic bias
Show answer & explanationAnswer & explanation
Correct answer: D. Algorithmic bias
The scenario describes the AI model making unfair decisions based on a proxy for a protected characteristic (postal code correlating with income/demographics), leading to discrimination. This is a classic example of algorithmic bias, where the model learns and perpetuates unfair patterns from data.
Why the other options are wrong
- A. AI privacy breach involves unauthorized data access, not discriminatory decisions.
- B. Lack of AI interpretability means not understanding *how* the decision was made, not that the decision itself is biased.
- C. AI robustness failure implies system instability or vulnerability, not inherent unfairness in decision-making.
Algorithmic Bias
Algorithmic bias occurs when an AI system produces systematically unfair, prejudiced, or discriminatory outcomes against certain groups, often stemming from biased training data or flawed model design.
- Can lead to discriminatory decisions in various domains.
- Often originates from historical biases in training data.
- Requires careful data collection, model evaluation, and mitigation strategies.
Memory trick: Bias: AI's blind spot.