I like the idea of alignments in general, but I think it's way too overused in D&D. It's a very useful tool for classic high fantasy campaigns where the conflict between forces of good and evil (and/or law and chaos, if you prefer that) is a major factor, but outside of that it's mostly a stumbling block for a lot of players.
Generally, unless you're actually using effects that really care about alignment (a divine plague that selectively strikes down evildoers, a holy artifact that can only be used by the pure of heart, and so forth), I find it's best to ignore it. However, when you are using such effects, it's a lot easier to have alignment as a shorthand for how a character interacts with it than to evaluate each one on a case-by-case basis.
Also, I find that a lot of the perceived issues with alignment can be fixed by viewing it as a loose coordinate system rather than a fixed set of discrete categories. Imagine plotting your character's general outlook and tendencies with respect to each alignment axis on a two-dimensional graph, then see where you wind up. Two characters might be in the same general region of the graph (have the same alignment), but still have significant differences in how they view alignment-relevant issues.
Generally, unless you're actually using effects that really care about alignment (a divine plague that selectively strikes down evildoers, a holy artifact that can only be used by the pure of heart, and so forth), I find it's best to ignore it. However, when you are using such effects, it's a lot easier to have alignment as a shorthand for how a character interacts with it than to evaluate each one on a case-by-case basis.
Also, I find that a lot of the perceived issues with alignment can be fixed by viewing it as a loose coordinate system rather than a fixed set of discrete categories. Imagine plotting your character's general outlook and tendencies with respect to each alignment axis on a two-dimensional graph, then see where you wind up. Two characters might be in the same general region of the graph (have the same alignment), but still have significant differences in how they view alignment-relevant issues.