Abstract
General context and questions
Adolescents and young adults typically use their smartphone several hours a day. Although there are concerns about how such behaviour might affect their well-being, the popularity of these powerful devices also opens novel opportunities for monitoring well-being in daily life. If successful, monitoring well-being in daily life provides novel opportunities to develop future interventions that provide personalized support to individuals at the moment they require it (just-in-time adaptive interventions). Taking an interdisciplinary approach with insights from communication, computational, and psychological science, this dissertation investigated the relation between smartphone app use and well-being and developed machine learning models to estimate an individual’s well-being based on how they interact with their smartphone.
To elucidate the relation between smartphone trace data and well-being and to contribute to the development of technologies for monitoring well-being in future clinical practice, this dissertation addressed two overarching questions:
RQ1: Can we find empirical support for theoretically motivated relations between smartphone trace data and well-being in individuals?
RQ2: Can we use smartphone trace data to monitor well-being in individuals?
Aims
The first aim of this dissertation was to quantify the relation between the collected smartphone trace data and momentary well-being at the sample level, but also for each individual, following recent conceptual insights and empirical findings in psychological, communication, and computational science. A strength of this personalized (or idiographic) approach is that it allows us to capture how individuals might differ in how smartphone app use is related to their well-being. Considering such interindividual differences is important to determine if some individuals might potentially benefit from spending more time on their smartphone apps whereas others do not or even experience adverse effects.
The second aim of this dissertation was to develop models for monitoring well-being in daily life. The present work pursued this transdisciplinary aim by taking a machine learning approach and evaluating to what extent we might estimate an individual’s well-being based on their smartphone trace data. If such traces can be used for this purpose by helping to pinpoint when individuals are unwell, they might be a useful data source for developing future interventions that provide personalized support to individuals at the moment they require it (just-in-time adaptive interventions). With this aim, the dissertation follows current developments in psychoinformatics and psychiatry, where much research resources are invested in using smartphone traces and similar data (obtained with smartphone sensors and wearables) to develop technologies for detecting whether an individual is currently unwell or will be in the future.
Data collection and analysis
This work combined novel data collection techniques (digital phenotyping and experience sampling methodology) for measuring smartphone use and well-being in the daily lives of 247 student participants. For a period up to four months, a dedicated application installed on participants’ smartphones collected smartphone trace data. In the same time period, participants completed a brief smartphone-based well-being survey five times a day (for 30 days in the first month and 30 days in the fourth month; up to 300 assessments in total). At each measurement, this survey comprised questions about the participants’ momentary level of procrastination, stress, and fatigue, while sleep duration was measured in the morning.
Taking a time-series and machine learning approach to analysing these data, I provide the following contributions: Chapter 2 investigates the person-specific relation between passively logged usage of different application types and momentary subjective procrastination, Chapter 3 develops machine learning methodology to estimate sleep duration using smartphone trace data, Chapter 4 combines machine learning and explainable artificial intelligence to discover smartphone-tracked digital markers of momentary subjective stress, Chapter 5 uses a personalized machine learning approach to evaluate if smartphone trace data contains behavioral signs of fatigue. Collectively, these empirical studies provide preliminary answers to the overarching questions of this dissertation.
Summary of results
With respect to the theoretically motivated relations between smartphone trace data and wellbeing (RQ1), we found that different patterns in smartphone trace data, from time spent on social network, messenger, video, and game applications to smartphone-tracked sleep proxies, are related to well-being in individuals. The strength and nature of this relation depends on the individual and app usage pattern under consideration. The relation between smartphone app use patterns and well-being is limited in most individuals, but relatively strong in a minority. Whereas some individuals might benefit from using specific app types, others might experience decreases in well-being when spending more time on these apps.
With respect to the question whether we might use smartphone trace data to monitor well-being in individuals (RQ2), we found that smartphone trace data might be useful for this purpose in some individuals and to some extent. They appear most relevant in the context of sleep monitoring (Chapter 3) and have the potential to be included as one of several data sources for monitoring momentary procrastination (Chapter 2), stress (Chapter 4), and fatigue (Chapter 5) in daily life.
Outlook
Future interdisciplinary research is needed to investigate whether the relationship between smartphone use and well-being depends on the nature of the activities performed on these devices, the content they present, and the context in which they are used. Answering these questions is essential to unravel the complex puzzle of developing technologies for monitoring well-being in daily life.
Adolescents and young adults typically use their smartphone several hours a day. Although there are concerns about how such behaviour might affect their well-being, the popularity of these powerful devices also opens novel opportunities for monitoring well-being in daily life. If successful, monitoring well-being in daily life provides novel opportunities to develop future interventions that provide personalized support to individuals at the moment they require it (just-in-time adaptive interventions). Taking an interdisciplinary approach with insights from communication, computational, and psychological science, this dissertation investigated the relation between smartphone app use and well-being and developed machine learning models to estimate an individual’s well-being based on how they interact with their smartphone.
To elucidate the relation between smartphone trace data and well-being and to contribute to the development of technologies for monitoring well-being in future clinical practice, this dissertation addressed two overarching questions:
RQ1: Can we find empirical support for theoretically motivated relations between smartphone trace data and well-being in individuals?
RQ2: Can we use smartphone trace data to monitor well-being in individuals?
Aims
The first aim of this dissertation was to quantify the relation between the collected smartphone trace data and momentary well-being at the sample level, but also for each individual, following recent conceptual insights and empirical findings in psychological, communication, and computational science. A strength of this personalized (or idiographic) approach is that it allows us to capture how individuals might differ in how smartphone app use is related to their well-being. Considering such interindividual differences is important to determine if some individuals might potentially benefit from spending more time on their smartphone apps whereas others do not or even experience adverse effects.
The second aim of this dissertation was to develop models for monitoring well-being in daily life. The present work pursued this transdisciplinary aim by taking a machine learning approach and evaluating to what extent we might estimate an individual’s well-being based on their smartphone trace data. If such traces can be used for this purpose by helping to pinpoint when individuals are unwell, they might be a useful data source for developing future interventions that provide personalized support to individuals at the moment they require it (just-in-time adaptive interventions). With this aim, the dissertation follows current developments in psychoinformatics and psychiatry, where much research resources are invested in using smartphone traces and similar data (obtained with smartphone sensors and wearables) to develop technologies for detecting whether an individual is currently unwell or will be in the future.
Data collection and analysis
This work combined novel data collection techniques (digital phenotyping and experience sampling methodology) for measuring smartphone use and well-being in the daily lives of 247 student participants. For a period up to four months, a dedicated application installed on participants’ smartphones collected smartphone trace data. In the same time period, participants completed a brief smartphone-based well-being survey five times a day (for 30 days in the first month and 30 days in the fourth month; up to 300 assessments in total). At each measurement, this survey comprised questions about the participants’ momentary level of procrastination, stress, and fatigue, while sleep duration was measured in the morning.
Taking a time-series and machine learning approach to analysing these data, I provide the following contributions: Chapter 2 investigates the person-specific relation between passively logged usage of different application types and momentary subjective procrastination, Chapter 3 develops machine learning methodology to estimate sleep duration using smartphone trace data, Chapter 4 combines machine learning and explainable artificial intelligence to discover smartphone-tracked digital markers of momentary subjective stress, Chapter 5 uses a personalized machine learning approach to evaluate if smartphone trace data contains behavioral signs of fatigue. Collectively, these empirical studies provide preliminary answers to the overarching questions of this dissertation.
Summary of results
With respect to the theoretically motivated relations between smartphone trace data and wellbeing (RQ1), we found that different patterns in smartphone trace data, from time spent on social network, messenger, video, and game applications to smartphone-tracked sleep proxies, are related to well-being in individuals. The strength and nature of this relation depends on the individual and app usage pattern under consideration. The relation between smartphone app use patterns and well-being is limited in most individuals, but relatively strong in a minority. Whereas some individuals might benefit from using specific app types, others might experience decreases in well-being when spending more time on these apps.
With respect to the question whether we might use smartphone trace data to monitor well-being in individuals (RQ2), we found that smartphone trace data might be useful for this purpose in some individuals and to some extent. They appear most relevant in the context of sleep monitoring (Chapter 3) and have the potential to be included as one of several data sources for monitoring momentary procrastination (Chapter 2), stress (Chapter 4), and fatigue (Chapter 5) in daily life.
Outlook
Future interdisciplinary research is needed to investigate whether the relationship between smartphone use and well-being depends on the nature of the activities performed on these devices, the content they present, and the context in which they are used. Answering these questions is essential to unravel the complex puzzle of developing technologies for monitoring well-being in daily life.
Original language | English |
---|---|
Qualification | Doctor of Philosophy |
Awarding Institution |
|
Supervisors/Advisors |
|
Award date | 4 Jul 2023 |
Publisher | |
Print ISBNs | 978-94-6483-111-5 |
Publication status | Published - 4 Jul 2023 |
Keywords
- Smartphones
- Procrastination
- Monitoring Sleep
- Stress
- Fatigue