Facial Indicators of Extraversion and Neuroticism: A Computational Approach

Measures of behavioural expressions to examine personality traits (like extraversion, neuroticism and agreeableness) of an encoder in face-to-face interaction are limited in social sciences research; subjective measures are available in psychometric literature in abundance, though. Efforts have been made in the field of artificial intelligence to detect facial expressions (a state marker of personality) of emotion of an encoder during dyadic interaction. Bio-algorithms have been developed accordingly to automate the decoding ability of the perceiver. Although detection of facial expressions is possible today, it has been found that emotion identification including transitory emotion expressions is also possible using machine learning or deep learning paradigms. Although in recent literature a number of works have been done on big 5 personality traits and facial attributes, a majority of them use static images. Moreover, the focus is more on automatic training using pixel properties. Work that focusses on identification of facial regions of activates or emphasized the movement of action units (AUs) are hardly available. However, such a method, especially using two-dimensional images, has significant implication in terms of behavioural sciences, since feature changes and their implications can be identified and correlated with other behavioural findings. Based on this gap area the present study attempts to develop algorithms to analyse the characteristics of eight facial regions from facial landmarks using unsupervised learning in order to identify patterns of activities for extraversion and neuroticism. An exploratory analysis of the patterns for the above traits using a database of IIT students suggested distinctive patterns for neuroticism and extraversion. Future research would explore the other traits of big 5 personality and attempt prediction on the basis of findings.