Basic Protocol: Smartphone Sensing Panel Study

Short summary: In this project we will investigate and compare the predictiveness of personality traits and objective sensing data from smartphones for life outcomes. Predicting life outcomes from Mobile Sensing Data Personality traits are generally described as relatively stable patterns of thought, feelings, and behaviors that are relevant in many parts of life. Further, personality traits are important because they predict important life outcomes (Ozer & Benet-Martínez, 2006, Roberts, Kuncel, Shiner, Caspi & Goldberg, 2007 Soto, 2019). Hence, knowing someone's personality is extremely useful to assess whether a person would be open to discuss ideas over a coffee, reliable to work with, good to have as a friend, or good to party with. While these findings are impressive, they are somewhat limited because they entirely rely on self-reported data from questionnaires and experience sampling and on in-sample correlations or regression modeling. The fundamental problems of self-reports have been known for a long time, especially with regard to behavioral data (Ellis, Davidson, Shaw & Geyer, 2019, Furr, 2009, Gosling, John, Craik & Robins, 1998, Paulhus & Vazire, 2007). Until recently, it was extremely difficult to collect large amounts of objective data on how people actually behave. However, with the advent of new computing technologies in the last two decades, researchers now have the tools to collect and analyze data from objective quantifications of individual differences to determine the role psychological traits and states play for life outcomes (Boyd, Pasca & Lanning, 2020). Especially promising are behavioral and situational data that can be gathered from smartphones (Harari, 2015, 2016, Miller, 2012). Further, it has been shown that these data are both associated with and predictive of personality traits at the factor and facet level (Harari, et al., 2019, Stachl et al., 2017, 2019). The richness, unobtrusiveness, objectivity and fine granularity of these data increasingly raises the question of whether self-reports should still be considered the gold standard or ground truth for the quantification of personality traits and individual differences (Boyd et al., 2020, Boyd & Pennebaker, 2017). In particular this question is relevant if prediction rather than explanation is the primary goal. The predictiveness of personality traits (for behavior and life outcomes alike) has been frequently praised in the literature (Ozer & Benet-Martínez, 2006, Roberts, Kuncel, Shiner, Caspi & Goldberg, 2007 Soto, 2019). However, predictive modeling (i.e., out-of-sample model evaluation) has rarely been used in past research to evaluate this claim. While, similar in-sample associations might be observable across replicating studies, it will be more interesting to see if models for the prediction of life outcomes can be created based on self-reports of personality and in-vivo behaviors and how well the models generalize beyond individual samples (Soto, in press). This motivates us to compare how well state of the art self-report measures of psychological traits (Big Five personality traits), and sensing data from smartphones can be associated with, and predict life outcomes.

Individual differences in cognitive abilities are known to predict various important life outcomes, making their study a critical area of interest for practitioners and researchers alike. While most research studied cognitive abilities within laboratory or achievement contexts, different lines of research investigated their role in everyday life, repeatedly linking them to our everyday behavior. However, as prior work mainly relied on reported behavior or simulated tasks, the relationship between cognitive abilities and objective behavior in everyday life remains unclear. The recent adaption of smartphone sensing and computational methods in psychology has demonstrated the potential of studying individual differences in real- world settings. In this fashion, the present study leverages digital footprints from everyday smartphone usage to investigate how fluid intelligence, one of the most central cognitive abilities within the Cattell-Horn-Carroll Theory (CHC; McGrew, 2009), is related to objective behavior in everyday life. More specifically, by means of a machine learning approach, we investigate (1) to what extent behavioral patterns in everyday smartphone usage predict fluid intelligence and (2) which behavioral patterns are most important for these predictions. For this purpose, we drew on existing literature to derive a comprehensive overview of behavioral correlates of fluid intelligence in everyday life capturable via logs of everyday smartphone usage. Translating these findings into features of multimodal smartphone usage data (e.g., phone usage duration, app installations, music consumption, typing patterns), we created a list of sensing features that correspond to the theory-based behavioral correlates and are described in this preregistration protocol. Using cross-validation, we will train linear and non-linear machine learning models (e.g., Elastic Net, Random Forest) based on these features and determine their predictiveness for participants’ composite scores of a fluid intelligence test. By means of interpretable machine learning techniques, we will examine which single features and feature groups contribute most to the predictive performance of these models.

The expression and recognition of emotions (i.e., short-lived and directed representations of affective states) through the acoustic properties of speech is a unique feature of human communication (Weninger et al., 2013). Researchers have identified acoustic features, which are predictable of affective states, and emotion detecting algorithms have been developed (Schuller, 2018). However, most studies used speech data produced by actors, who had instructions to act out a given emotion, or speech samples labelled by raters, who were instructed to add affective labels to recorded utterances (e.g., from TV shows). Both, enacted and labelled speech, come with multiple downsides since these approaches assess expressed affect rather than the experience of actual affective states through voice. In this work, we want to investigate if we can predict in-situ self-reported affective states from objective voice parameters collected with smartphones in everyday life. Further, we want to explore which acoustic features are most predictive for the prediction of the experience of affective states. Finally, we want to analyze how the affective quality of instructed spoken language (e.g., a sentence with negative affective valence) translates into objective markers in the acoustic signal, which then in turn could alter the predictions in our models.

We present the dataset for the article "Keep on scrolling? Using intensive longitudinal smartphone sensing data to assess how everyday smartphone usage behaviors are related to well-being". The data were collected as part of the Smartphone Sensing Panel Study and comprise several dataset parts, as we replicated our analysis for two different 14-day measurement periods (A and B). At the macro level, we aggregated different measures of smartphone use (measured by mobile sensing) over 14 days and examined their associations with global survey-based measures of well-being (Flourishing, Satisfaction WIth Life, Positive Activation, Negative Activation, Valence; Dataset A: N = 236, Dataset B: N = 305). At the micro level, we aggregated various measures of smartphone use (measured via mobile sensing) over 60-minute windows before asking participants about their current mood using experience sampling questionnaires (Dataset A: N = 378, n = 5775; Dataset B: N = 534, n = 7287). In our supplementary analysis, we also aggregated the smartphone usage data for 15-minute windows to analyse social and non-social situations. Demographic variables (age, gender, education) that were not used for the data analyses were removed for privacy reasons, but can be provided upon request. The datasets are documented by a comprehensive accompanying codebook. Additional materials (e.g., preprocessing and analysis code) can also be found at https://osf.io/ckwge/ Further details on the variables provided and the associated study procedures can be found in the journal article: große Deters, F., & Schoedel, R. (2024). Keep on scrolling? Using intensive longitudinal smartphone sensing data to assess how everyday smartphone usage behaviors are related to well-being, Computers in Human Behavior, 150, 107977, https://doi.org/10.1016/j.chb.2023.107977

Right-wing Authoritarianism (RWA) is on the rise globally. As a result, researchers are trying to understand how this defining aspect of 20th century history is manifesting in the 21st century digital era. Individual differences in RWA have been studied extensively from a theoretical standpoint. In addition, a great deal of empirical research has examined the situational antecedents (e.g., threat) and attitudinal consequences (e.g., prejudice) of authoritarianism. However, this theoretical and empirical work has failed to paint a holistic picture of authoritarians’ behaviors in daily life. Here, digital traces from smartphone use, which have informed the study of individual differences in other domains, represent a promising means to investigate the new authoritarians in their daily lives. Given the importance of understanding this phenomenon in the current geopolitical context, our study will create a theoretically-informed profile of everyday behaviors related to authoritarianism in the digital era. To this end, we drew on the literature to derive a comprehensive overview of empirical reports on behavioral indicators of authoritarianism. We then translated these findings into behavioral features (organized into five interrelated theoretical attributes) which can be captured using data collected from smartphone sensors and logs (e.g., communication, app-use, mobility, music/podcast consumption). Where possible, we plan to enrich sensed behavioral data with data from other sources to ensure that our features reflect the theoretical claims about authoritarianism as closely as possible. Lastly, we will use cross-validated machine learning models (i.e., Elastic Net and Random Forest) to determine whether we can predict self-reported authoritarianism from these behavioral features, using data from a representative sample of 749 participants who were tracked continuously for up to six months.

Smartphone usage might interfere with and interrupt offline activities . These interruptions can be exogenous or endogenous . The resulting distraction and disengagement from the unmediated environment can pose a threat to affective well-being. So far, the focus in the literature on the effects of smartphone usage and well-being is on total time of usage or usage at specific times of the day (e.g., nighttime). However, we hypothesize that not necessarily longer periods of smartphone usage but the continuous small interruptions during daily life are detrimental to our affective well-being. We will combine smartphone sensing, experience sampling, and questionnaire data collected in the Smartphone Sensing Panel study (SSPS) to investigate this hypothesis on an inter- and intra-individual level. In this preregistration protocol, we specify diverse plausible operationalizations for both predictors and outcomes. We also preregister the multiverse of models we will build based on the variety of variables available.

This study is conducted as part of a master thesis at the Department of Psychology of the Ludwig Maximilian University Munich. It investigates whether behavioral and situational data collected via smartphone sensing in daily life can predict individuals' psychological situation. For this purpose, this study applies a machine learning approach to predict individuals’ in situ ratings of perceived situational characteristics (DIAMONDS; Rauthmann et al., 2014) based on smartphone sensing data. Note that all independent variables (or features) defined in our preregistration protocol are also included in another study aimed for publication. All data used in this study is retrieved from the Smartphone Sensing Panel Study (SSPS; Basic Protocol of the SSPS is available under: http://dx.doi.org/10.23668/psycharchives.2901

Analyzing language offers a unique window into the inner workings of the human mind. Particularly, the methodological advancements in text processing in recent years and the ubiquity of textual digital trace data have generated opportunities to investigate psychological constructs, such as affective states, through language in the form of text. In this manner, prior studies have predicted affective states from text data, for example social media posts. Most prior studies relied on human annotations or on establishing sentiment through lexica (e.g., LIWC, VADER) to infer users’ affect from social media language samples (e.g., Facebook status updates). However, affective word usage and human judges’ rating conceptually differ from one’s subjective affect experience. Here, phone-level data collection methods offer a promising opportunity to passively log textual data across communication channels (public and private contexts) through the smartphone’s keyboard and couple the data with in-situ self-reports on one’s affect through experience sampling. In this work, we want to investigate (in-sample) associations of in-situ self-reported affective states with language features logged with smartphones in everyday life and if these features allow for the (out-of-sample) prediction of between-person differences and within-person fluctuations in affect experience. Further, we want to investigate which language features are most predictive of subjective affect experience. Finally, we want to analyze what the optimal time window is for text analysis (and corresponding amount of text data) around the timestamp of the affective state in question and how affect experience is revealed in different contexts (e.g., public posting vs. private messaging).

Momentary experiences of positive and negative emotionality are core components of well-being and performance. This study investigates whether passively sensed smartphone data can be used to recognize individuals’ mood (i.e. Valence and Arousal (Russell, 1980)) based on their smartphone sensing data. The exploratory analysis uses data generated from N = 453 participants in a two-week experience sampling wave which was part of the Smartphone Sensing Panel Study (SSPS; Schödel & Oldemeier, 2020). Different cross-validated machine learning algorithms are compared to predict participants’ current mood given a variety of situational and behavioral variables, reflected by different smartphone sensing modalities. Moreover, the impact of different time perspectives (i.e. daily versus hourly) on the predictive performance is investigated.

We present the dataset for the article "Snapshots of Daily Life: Situations Investigated Through the Lens of Smartphone Sensing." The data were collected as part of the Smartphone Sensing Panel Study and include 9,790 situational snapshots (observations) from N = 455 participants collected over 14 days of daily life using mobile sensing and experience sampling. Specifically, Dataset 1 is an aggregated mobile sensing dataset with 1,365 cues (including variables extracted from GPS, phone, app, activity logs, etc.) and experience sampling on situational awareness and affective valence. Dataset 2 contains the Big Five as person variables. Demographic and technical variables (age, gender, education, manufacturer, and Android version of the smartphone) that were not used for the data analyses were removed for privacy reasons, but can be made available upon request. The datasets are documented by a comprehensive accompanying codebook. Additional materials (e.g., preprocessing and analysis code) can also be found at https://osf.io/b7krz/. Further details on the variables provided and the associated study procedures can be found in the journal article: Schoedel, R., Kunz, F., Bergmann, M., Bemmann, F., Bühner, M., & Sust, L. (Accepted). Snapshots of Daily Life: Situations Investigated Through the Lens of Smartphone Sensing. Accepted for Publication in: Journal of Personality and Social Psychology.