It is an artefact of how we scan things visually. Most of us read top-to-bottom/left-to-right and habitually look at pictures with that bias, so when the brain fills in the missing height detail for a 2D display using the light/shadow indicators it defaults to assuming the light source is in the top left. The affect designed for us doesn't work as well for those whose primary written forms follow a right-to-left structure. Obviously, as you mention, the light source being assumed to be high comes from a much older source as our main natural lights are up high.