An illicit theory of costly signaling — LessWrong