Lean toward: moral realism. My leanings are semantic. There is not one unified object or semantic value that all members of our linguistic community intend by 'moral.' So we must either choose one of the semantic values in advance, or leave all moral statements underdetermined. I think choosing a semantic value that naturalizes morality (e.g., 'morality is behaving in accord with a decision procedure that optimizes for the overall preference satisfaction of all preference-bearers') is much more useful and conducive to our goals than choosing one that forfeits all moral convictions to the Dark Side. If we want to win, we have to convince people in general to move toward a LessWrongier perspective; and convincing people without making any appeal to words like 'good,' 'bad,' 'right,' 'wrong,' 'moral,' or 'immoral' is a hell of a handicap.