Personality Constructs Predictions Beyond FFM/Big5: A Digital Phenotyping-Based Exploration

Area of Science:

• Psychology
• Computer Science
• Data Science

Background:

• Digital phenotyping leverages smartphone data for continuous, in situ behavioral data collection, extending traditional Experience Sampling Methods (ESMs).
• This research moves beyond the Five-Factor Model (FFM)/Big Five personality constructs to encompass a broader range of 59 traits and types from 16 distinct personality theories, including temperament and personal values.

Purpose of the Study:

• To investigate the efficacy of digital phenotyping via smartphone data in predicting a wide array of personality constructs.
• To compare hypothesis-testing and machine learning approaches for analyzing digital footprint data in personality research.
• To explore the potential of digital phenotyping for remote psychological assessment and People Analytics.

Main Methods:

• Collected digital footprints from 104 participants' smartphones over 7-10 days.
• Employed both deductive (hypothesis-testing) and inductive (machine learning) analytical methods.
• Utilized machine learning models including Gradient Boosted Trees, Decision Tree, Random Forest, and Support Vector Machine.

Main Results:

• Successfully predicted 4 out of 16 personality constructs (25%), including Adult Attachment, FFM/Big5, Distress Tolerance, and Creativity, with prediction accuracy (r) ranging from 0.034 to 0.53.
• Achieved an overall prediction success rate of 37.29% for 22 out of 59 individual traits and types.
• Identified Gradient Boosted Trees as the most effective machine learning model, particularly for communication-related data features.

Conclusions:

• Digital phenotyping of smartphone data demonstrates significant potential for advancing remote personality psychology research.
• The findings highlight the applicability of digital phenotyping in People Analytics and other interdisciplinary fields.
• This approach offers a scalable and objective method for understanding individual differences in personality.