A skewed sample occurs when historical data is biased, unrepresentative, or imbalanced, leading to discriminatory AI outcomes. In this case, the AI system was trained on 20 years of national data, which likely contained historical biases against minorities in drug and alcohol disorder treatments.

Since past law enforcement and healthcare policies have disproportionately impacted minority groups, the AI inherited these biases and began making unfair, discriminatory predictions.

Algorithmic Bias & AI Ethics (ACM/IEEE Guidelines) – AI models must be trained on diverse and representative data.

Fairness in Machine Learning (FAT/ML Framework) – Preventing bias in AI requires identifying and correcting skewed data.

Historical Bias & Systemic Discrimination – Training AI on biased past records leads to unethical and discriminatory outcomes.

Deontological Ethics (Fair AI Decision-Making) – AI developers must ensure unbiased, ethical, and fair AI outputs.

Relevant Ethical References in Technology:Thus, the correct answer is B. Skewed sample, as the AI model was trained on biased historical data, leading to discrimination.