Daniel Preotiuc

Abstract

The continuous increase in user-generated content allows researchers to study data in a more complex setting, including information about its authorship. I will provide an overview of my research on using computational methods to automatically analyse the public information generated by users on social media and predict with high accuracy various demographic and psychological traits, with use cases including socio-economic factors, personality, depression in medical records and political ideology. I will demonstrate how we can uncover relationships between the trait of interest and information such as concepts, topics or emotions expressed through text, image content and choice or usage patterns. Further, I will present work on quantifying biases in how humans perceive the traits of others from their posts. These studies offer data-driven insights into human and group behaviour and can be used to suggest causal hypotheses or interventions. I will end by highlighting extensions in this area of research and discuss algorithmic implications.

Bio

Daniel Preotiuc-Pietro is a senior research scientist at Bloomberg. His research addresses real-world problems in the social sciences through computational analysis of large scale user-generated data, usually from social media. Previously, Daniel completed his postdoc at the University of Pennsylvania and obtained a PhD in Natural Language Processing and Machine Learning at the University of Sheffield. His research has been featured in popular press outlets, including the Washington Post, New Scientist, Scientific American and the BBC.