Underfitting: A common problem in machine learning where a model performs poorly on both the training and new, unseen data. It occurs when a model is too simple and is not able to learn the underlying pattern in the data.