Ambient Decision Theory — LessWrong