# Minimum image convention and the atom “seeing” itself

The minimum image convention in molecular dynamics can be state as: A particle in the central square must not interact with one of its neighbors in the central square and an image of this neighbor in one of the replicas at the same time but only with the closer one.

My question is: what is the basis for this convention? In other words, what is the problem with a central atom seeing another atom from both the image and the central box?
If I look at the simulation box as just part of an infinite bulk material, then I should be able to take the cutoff radius as large as I like. No matter how many images of an atom are seen.