Personality Computing With Naturalistic Music Listening Behavior: Comparing Audio and Lyrics Preferences

Abstract

Psychologists have long theorized that people use music to create auditory environments matching their personality traits. While there is initial evidence relating self-reported musical style preferences to the Big Five dimensions, little is known about day-to-day music listening behavior and the intrinsic attributes of music that give rise to personality patterns. The present study (N = 330) proposes a personality computing approach to fill these gaps with new insights from ecologically valid music listening records from smartphones. We provided a holistic account of individual differences in music listening by integrating everyday preferences for various musical attributes with habitual listening behaviors. More specifically, we quantified music preferences at fine granularity via technical audio features from Spotify and via lyrical attributes obtained through natural language processing. Using machine learning algorithms, these behavioral variables served to predict Big Five personality on domain and facet level. Our out-of-sample prediction accuracies revealed that the Openness dimension was most strongly related to music listening, while several other traits (most notably Conscientiousness facets) also showed moderate effects. Thereby, preferences for audio and lyrics characteristics were distinctly predictive of personality, hinting at the incremental value of both musical components. Furthermore, variable importance metrics displayed generally trait-congruent relationships between personality outcomes and music listening behaviors prompting us to discuss possible mechanisms underlying these interactions. Overall, our study contributes to the development of a detailed cumulative theory on music listening in personality science, which may be extended in numerous ways in future studies leveraging the computational framework proposed here.

Snapshots of daily life: Situations investigated through the lens of smartphone sensing.

“…We found that the successful prediction of situation characteristics relied on a broad range of different smartphone-sensed cues ranging between 22 and 46 per DIAMONDS dimension (see Table 3). Thus, analogous to personality prediction research (Schoedel et al, 2018;Stachl et al, 2020;Sust et al, 2022), single cues contributed little information on their own, while their holistic constellation was much more informative.…”

“…The sensed data cover a wide range of modalities and all five groups of situation cues (Harari et al, 2015; Harari, Müller, & Gosling, 2020). For example, interactions can be inferred from call records (e.g., Harari, Müller, & Stachl, 2020; Servia-Rodríguez et al, 2017; Wang et al, 2016), objects from music listening records (e.g., Sust et al, 2022; Yang & Teng, 2015), (physical) activities from accelerometer sensors (e.g., Servia-Rodríguez et al, 2017; Wang et al, 2016; see, Ramanujam et al, 2021, for a review of human activity recognition), locations from GPS sensors (e.g., Canzian & Musolesi, 2015; Do & Gatica-Perez, 2014; Müller et al, 2020), and time frames from the timestamps of a given situation (Böhmer et al, 2011; Stachl et al, 2020).…”

Snapshots of Daily Life: Situations Investigated Through the Lens of Smartphone Sensing

“…The sensed data cover a wide range of modalities and map onto all five groups of situation cues (Harari et al, 2015;. For example, Interactions can be inferred from call records (e.g., Harari, Müller, Stachl, et al, 2020;Servia-Rodríguez et al, 2017;Wang et al, 2016), Objects from music listening records (e.g., Sust et al, 2022;Yang & Teng, 2015), (physical) Activities from accelerometer sensors (e.g., Servia-Rodríguez et al, 2017;Wang et al, 2016;s., Ramanujam et al (2021) for a review of human activity recognition), Locations from GPS sensors (e.g., Canzian & Musolesi, 2015;Do & Gatica-Perez, 2014;Müller et al, 2020), and Time frames from the timestamps of a given situation (Böhmer et al, 2011;Stachl et al, 2020). For a more detailed overview of smartphone sensors, please refer to Lane et al (2010) or our method section.…”

“…To describe these songs in terms of their intrinsic musical attributes, we enriched them with song-level variables provided by Spotify's Track API 5 . The resulting ten variables reflect the audio characteristics (e.g., danceability, energy, loudness) of songs listened to and are described in more detail in Sust et al (2022) and the Codebook in the online supplemental material.…”

Best Practices in Supervised Machine Learning: A Tutorial for Psychologists

“…Throughout the tutorial, we use the publicly available PhoneStudy behavioral patterns dataset, which has been used to predict human personality from smartphone usage data (Stachl, Au, et al, 2020). Subsets of these data have also been used in a number of other publications (Au et al, 2021;Harari et al, 2019;Schoedel et al, 2020Schoedel et al, , 2018Schuwerk et al, 2019;Stachl et al, 2017;Sust et al, 2022). The dataset contains self-reported questionnaire data of personality traits measured with the German Big Five Structure Inventory (BFSI; 5 factors and 30 facets, Arendasy et al, 2011), demographic variables (age, gender, education), and behavioral data from smartphone sensing (e.g., communication & social behavior, app-usage, music consumption, overall phone usage, day-nighttime activity).…”