If, on the contrary, we suffer from high variance, it means that our model is too complex for the data. In this case, we can only try to get more data or decrease the complexity. This would mean increasing k so that more neighbors would be taken into account, or removing some of the features.