Underfitting is a phenomenon in machine learning where a model is too simplistic to adequately represent the underlying structure of the data. This can happen when the model lacks sufficient complexity or when it has not been trained long enough. As a result, underfitted models fail to capture important patterns, leading to poor performance on both training and test datasets. Common indicators of underfitting include high bias and low accuracy in predictions. To mitigate underfitting, practitioners may increase model complexity, add relevant features, or employ more sophisticated algorithms that can better capture the intricacies of the data. Striking a balance between model complexity and performance is essential to achieving optimal results in machine learning tasks.