Age and gender in language, emoji, and emoticon usage in instant messages

Text is one of the most prevalent types digital data that people create as they go about their lives. Digital footprints people's language usage in social media posts were found to allow for inferences age and gender. However, even more potentially sensitive text from instant messaging services has remained largely uninvestigated. We analyze variations messages with regard individual differences gender by replicating extending methods used prior research on posts. Using a dataset 309,229 WhatsApp 226 volunteers, we identify unique age- gender-linked variations. use cross-validated machine learning algorithms predict volunteers' (MAEMd = 3.95, rMd 0.81, R2Md 0.49) (AccuracyMd 85.7%, F1Md 0.67, AUCMd .82) significantly above baseline-levels predictive features. discuss implications psycholinguistic theory, present opportunities application author profiling, suggest methodological approaches making predictions small sets. Given recent trend towards dominant private increasingly weaker user protection, highlight rising threats privacy rights messaging.