The Alignment Paradox: Why Transparency Can Breed Deception — LessWrong