Automatic Assessment of Personality Traits Using Non-verbal Cues in a Saudi Sample

摘要

Different factors shape individuals' personality, where their interaction with others in certain situations could reveal their personal characteristics. Studies have explored the influence of culture in forming personalties, where differences in behaviours are observed. The advancement in communication technologies has opened the world and increased the cultural diversity. Therefore, understanding individual personalities is crucial for the enhancing the effectiveness in communication and for the development of an interconnected world. Such an understanding not only would guarantee smooth group interaction in workplace, education, and social environments, but also would allow for better resource utilization and role allocation for group members. Moreover, with the emergence of HCI technologies and affective computing, automation of personality assessment using non-verbal cues seems feasible. Acknowledging the differences in personality traits between cultures, several studies have analysed such traits clusters in different countries. However, given the unique culture of Arabs in general and Saudi Arabian in particular, personality traits distribution is yet to be investigated. This research investigates two aspects: (1) the distribution of personality types of individuals living in Saudi Arabia compared to other countries, and (2) the feasibility of automatically classifying personality types by analysing non-verbal cues during an interaction setting. To accomplish the first part of the this work, we used the big-five personality assessment survey, where a total of 232 individuals have responded. The results showed a slight difference in the personality assessment of individuals living in Saudi Arabia compare to other cultures. For the second part, we conduced physical interviews with eight subjects where their body actions are recorded. Several non-verbal features were extracted from the body movement (e.g. touching face) and used for automatic classification. The results are generally reasonable, where the accuracy on average was 67% using Support Vector Machines. The slight differences in the personality types from this study results compared suggest the uniqueness of Arab culture in general and Saudi culture in particular. Moreover, the automatic assessment of personality types using body language demonstrate a potential success. Linking the two aspects of personality distribution and automatic assessment of personality, could increase the reliability and accuracy of the results.