In contemporary digital culture, everyday life is permeated by computer-mediated activities. As a byproduct, a reservoir of digital trace data is generated. Can this data be leveraged to infer one’s personality?

Given the importance of personality prediction in several compelling applications ranging from mental health monitoring to occupational proficiency, in recent years there has been an increasing interest in developing new methods to predict personality traits automatically, by relying on digital traces. However, previous studies have focused on a single specific modality such as language patterns or profile pictures; never before has heterogeneous digital trace data been analyzed holistically.

By relying on the use of avatars, emoticons, language, and responsive patterns in social media as documenting slices of users’ everyday interactions, the authors demonstrate that it is possible to infer user personality traits in an effective and efficient way. Additionally, they provide useful insights about the correlation between different aspects of heterogeneous digital footprints and user personality. The proposed approach consists of first extracting semantic representations from each personality cue through feature engineering, then in extracting classification scores for each domain separately, and finally in using the set of resulting scores as input of a stacked generalization-based ensemble method.

Performances obtained with large-scale experiments compare favorably with respect to state-of-the-art methodologies such as MyPersonality and Personality Insights, also demonstrating the ability of the system to accurately distinguish users with extremely high scores on a given single personality trait from those with medium scores spread out over several traits. Furthermore, the authors tested the reliability of the proposed framework in a real-world scenario where a chatbot, called DiPsy, acts as a personalized digital psychologist able to predict the user’s mental status based on digital footprints, to revise the prediction results based on its understanding during natural language conversations, and eventually to conduct cognitive behavioral therapy.

Overall, this ambitious study shows that cyber-psychology is becoming a reality. Therefore, it can be a good read for researchers in the fields of social media analysis, multimedia applications, multimodal fusion, and social signal processing.