Social effects of algorithms that accurately identify human behaviour and traits — LessWrong