Latent variables are factors that cannot be observed directly but can be inferred.

The study revealed latent biases embedded in the machine learning training dataset.