A decision-support system is used across several organizational environments to inform outcomes that affect different...

This scenario describes a classic case of algorithmic bias rooted in historical data . The system is functioning correctly from a technical standpoint, but the training data reflects existing societal or structural inequalities , which are then reproduced in the model’s outputs.

Bias and fairness issues occur when:

Training data contains imbalances across demographic or population groups

Historical patterns encode discrimination or unequal access/opportunity

The model learns and perpetuates these patterns in predictions or decisions

This leads to systematic differences in outcomes , even without explicit errors in the algorithm.

Other options are not appropriate:

Overfitting relates to memorizing training data and poor generalization, not systemic group disparities

Data drift refers to changes in data distribution over time after deployment

Edge case failures involve rare or unusual scenarios, not consistent group-level differences

CAIPM governance principles emphasize that identifying bias requires understanding data provenance and historical context , not just model performance metrics.

Therefore, the correct answer is Bias and fairness issues , as it directly explains outcome disparities driven by structural imbalances in historical data.