Mean Absolute Error (MAE) vs Root Mean Squared Error (RMSE)

Hi all,

I’m quite lost in the use cases of MAE and RMSE as error metrics. In what circumstances would you use MAE as an error metric over RMSE?

I know theoretically MAE should be used for models where an error of 10 is only twice as bad as an error of 5, whereas RMSE should be used for models where an error of 10 is more than twice as bad. However, pragmatically, what would be an example of this? I tried searching all over the net but the only answer that I’ve found is the theoretical answer.

Thanks!