The Flexibility of Abstract Concepts — LessWrong