When I started my law vs. chaos alignment series I glossed over good and evil. I did this for 2 reasons, the first is that I was intentionally focusing on just the law and chaos axis on the alignment scale. The second reason was that I do feel that good vs. evil is much easier for players to wrap their heads around.
In both my initial post, and my post on moral relativism, I made some quick comments about good vs. evil being more easily definable. In particular in the moral relativism one I briefly touched on the concept that mechanically good vs. evil should be absolute. That's what I wanted to expand on a little more here.
The reason I said it needs to be absolute in the game mechanics is that in Dungeons and Dragons, and Pathfinder, there are game mechanics built around alignment. There are several spells and effects that only work on characters with a certain alignment. If you allow the mechanics to bend on what it means to be good or evil then it becomes far too easy for a character to write one alignment on their sheet but play another. I've seen this quite a bit with neutral characters, who by rights should really be evil, except for the fact that they're playing with the "good guys".
In a game system that has hard and fast rules around being good or evil I think there needs to be a well defined criteria for what it means to be good or evil. Otherwise you run the risk of arguments around alignment shifts or how somebody can have a certain alignment on their sheet (player or NPC) yet still perform a certain act.
I think it's a very worthwhile exercise for a GM to figure out exactly what that criteria will be and to spell it out for the players. Feel free to engage them in the discussion as well. Some food for thought that tend to come up a lot would be things like torture, killing defenseless foes, killing innocents, etc. Where things can really get hazy though is if you start applying this to a society. For instance, can a society that allows slavery be a good-aligned society?