Personality is Revealed During Weekends: Towards Data Minimisation for Smartphone Based Personality Classification
This work addresses the need for data minimisation to improve user engagement and privacy in mobile services, representing an incremental advance in personality modelling.
The study tackled the problem of reducing data collection duration for smartphone-based personality classification, achieving state-of-the-art accuracy of 66% to 71% for classifying five personality traits using only one or two weekends of data.
Previous literature has explored automatic personality modelling using smartphone data for its potential to personalise mobile services. Although passive modelling of personality removes the burden of completing lengthy questionnaires, the fact that such models typically require a few weeks or months of personal data can negatively impact user's engagement. In this study, we explore the feasibility of reducing the duration of data collection in the context of personality classification. We found that only one or two weekends can suffice for achieving state-of-the-art accuracy between 66% and 71% for classifying the five personality traits. These results provide lessons for practicing "data minimisation" - a key principle of privacy laws.