Predicting Personality from News Consumption

Matthew Edwards, Stephan Lewandowsky, Adam Sutton – University of Bristol

Personality information can be used as part of adversarial manipulation of an individual, making them more vulnerable to attacks on their reasoning process – whether the adversary here is a criminal phishing for card details or a marketing company trying to sell a product of dubious quality. Recent survey results show that people are particularly reluctant to receive targeted political messages, and especially so when this targeting takes place on the basis of information about their personality – a violation of their privacy that is leading to a perceived harm.

Existing work demonstrates that personality information about an individual can be predicted based on their online footprint, opening the door to targeted political manipulation. New PETs are now attempting to address this problem through better protection of online footprints. Our study aims to develop this area, building up an understanding of how personality can be predicted based on seemingly innocuous and hard-to-protect interactions such as reading and commenting online. This work will feed into the design of interventions that can detect when political message content is matched for consumption by particular personalities and inform users when material they are reading is suspiciously tailored to their own personality.