Training Data
The data used to teach a machine-learning model to make predictions. Its quality and representativeness directly affect fairness and accuracy.
Training data is the information used to teach a machine-learning model. The model learns patterns from this data and uses them to make predictions on new inputs. The quality, completeness, and representativeness of training data directly affect model accuracy and fairness.
Biased or unrepresentative training data is one of the most common sources of algorithmic discrimination. If the data underrepresents a protected class, overrepresents a particular geography, or reflects historical discrimination, the model will reproduce those patterns.
AI governance programs should document training data sources, cleaning steps, sampling methods, and known limitations. See our glossary entries on algorithmic bias, data lineage, and external consumer data.