vi. Data Quality, Bias & Model Risk

Data quality, bias, and model risk describe the limitations and potential failures of analytical and AI systems. Understanding these risks is essential to ensure that outputs are reliable, interpretable, and used appropriately in decision-making.

Bias

Systematic distortion in data or models that leads to skewed or unfair outcomes.

Data Drift

Changes in data patterns over time that reduce model accuracy or relevance.

Data Quality

The degree to which data is accurate, complete, consistent, and timely.

Error Propagation

The compounding effect of inaccuracies as data moves through analytical processes.

Explainability

The ability to understand how a model produces its outputs.

Model Risk

The potential for losses or poor decisions resulting from incorrect or misapplied models.

Outlier

A data point that deviates significantly from expected patterns and may distort analysis.

Sampling Bias

Bias introduced when training data does not represent the broader population or use case.

Validation Failure

A condition where a model performs poorly when tested against new or real-world data.

Use Limitation

Constraints defining when and how a model’s outputs should be applied.