Automated personality inference from social media data

Abstract

This thesis provides an overview of different methods used to perform what is called automated psychological trait inference and demonstrates the conflict between improving marketing communications performance and the ethical questions of using such methods. These methods comprise of a mix of social media data types, new technology, as well as personality assessment frameworks, of which the Five-Factor Model (FFM) of personality trait assessment (or “Big Five”) in particular. Methods of psychological trait inference are applied in a modern world of text and image mining allowing the discovery and analysis of new data points by training self-learning prediction algorithms. The methods are placed in a broader frame: a modern economic construction that is built on the commercial and strategic value of data. The study on the methods is limited to two social media platforms: Twitter and Facebook. Practical implications are provided along with ethical considerations.