Why should we take filters of dimensions 3x3 and not 2*2

In the edge detection, would it not be better to use 22 filter rather than 3x3? If we take 22, then we can exactly identify the placement of the edge, in 33, there are two columns with same max values in the output metrics, where as if we take 22, then only one column will have one column with max values.

Hi storm95,

Maybe this explanation clarifies.