Modeling personality vs. modeling personalidad: In-the-wild mobile data analysis in five countries suggests cultural impact on personality models
File(s)
Accepted version
OA Location
Author(s)
Type
Journal Article
Abstract
Sensor data collected from smartphones provides the possibility to passively infer a user’s personality traits. Such models can
be used to enable technology personalization, while contributing to our substantive understanding of how human behavior
manifests in daily life. A significant challenge in personality modeling involves improving the accuracy of personality
inferences, however, research has yet to assess and consider the cultural impact of users’ country of residence on model
replicability. We collected mobile sensing data and self-reported Big Five traits from 166 participants (54 women and 112
men) recruited in five different countries (UK, Spain, Colombia, Peru, and Chile) for 3 weeks. We developed machine learning
based personality models using culturally diverse datasets - representing different countries - and we show that such models
can achieve state-of-the-art accuracy when tested in new countries, ranging from 63% (Agreeableness) to 71% (Extraversion)
of classification accuracy. Our results indicate that using country-specific datasets can improve the classification accuracy
between 3% and 7% for Extraversion, Agreeableness, and Conscientiousness. We show that these findings hold regardless of
gender and age balance in the dataset. Interestingly, using gender- or age- balanced datasets as well as gender-separated
datasets improve trait prediction by up to 17%. We unpack differences in personality models across the five countries, highlight
the most predictive data categories (location, noise, unlocks, accelerometer), and provide takeaways to technologists and
social scientists interested in passive personality assessment.
be used to enable technology personalization, while contributing to our substantive understanding of how human behavior
manifests in daily life. A significant challenge in personality modeling involves improving the accuracy of personality
inferences, however, research has yet to assess and consider the cultural impact of users’ country of residence on model
replicability. We collected mobile sensing data and self-reported Big Five traits from 166 participants (54 women and 112
men) recruited in five different countries (UK, Spain, Colombia, Peru, and Chile) for 3 weeks. We developed machine learning
based personality models using culturally diverse datasets - representing different countries - and we show that such models
can achieve state-of-the-art accuracy when tested in new countries, ranging from 63% (Agreeableness) to 71% (Extraversion)
of classification accuracy. Our results indicate that using country-specific datasets can improve the classification accuracy
between 3% and 7% for Extraversion, Agreeableness, and Conscientiousness. We show that these findings hold regardless of
gender and age balance in the dataset. Interestingly, using gender- or age- balanced datasets as well as gender-separated
datasets improve trait prediction by up to 17%. We unpack differences in personality models across the five countries, highlight
the most predictive data categories (location, noise, unlocks, accelerometer), and provide takeaways to technologists and
social scientists interested in passive personality assessment.
Date Issued
2019-09-01
Date Acceptance
2019-08-15
Citation
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2019, 3 (3), pp.1-24
ISSN
2474-9567
Publisher
Association for Computing Machinery
Start Page
1
End Page
24
Journal / Book Title
Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies
Volume
3
Issue
3
Copyright Statement
© 2019 Copyright held by the owner/author(s). Publication rights licensed to ACM.
Identifier
https://dl.acm.org/citation.cfm?doid=3361560.3351246
Publication Status
Published
Article Number
88
Date Publish Online
2019-09-01