Gabrielle Liu created an algorithm that can be used by AI assistants to detect the risk for verbal and emotional abuse, by studying language.