Dr. Egon Dejonckheere
Egon does research in Emotion, Clinical Psychology and Abnormal Psychology. His most recent publication is 'The Bipolarity of Affect and Depressive Symptoms', featured in Journal of Personality and Social Psychology: Personality Processes and Individual Differences.
What is passive sensing?
Passive sensing is a research approach for measuring behavior, context, and physiological processes as they occur in people’s daily lives. Instead of asking individuals to report on their experiences, passive sensing collects data automatically from smartphones, smartwatches or other wearables, often continuously or at high frequency, in everyday environments. This enables researchers to capture real-world behavior with minimal participant burden and without relying on recall.
Passive sensing is part of the broader family of ambulatory assessment methods, which aim to capture real-life data outside the lab room. What distinguishes passive sensing within this framework is its reliance on device-based measurements rather than self-report, combining embedded sensors to approximate patterns of activity, mobility, social context, and physiology. In contrast to Ecological Momentary Assessment (EMA), Experience Sampling Methodology (ESM) or daily diary designs, it therefore does not require active participant input. Instead, it complements these approaches by capturing behavioral and contextual processes that may be difficult, impractical, or impossible to assess through self-report alone.
A helpful way to conceptualize passive sensing is as a continuous data layer underlying everyday life. Rather than sampling discrete moments through repeated assessments, passive sensing generates high-resolution streams of data over time. These data can be used not only to reconstruct patterns of behavior and context, but also to detect meaningful changes as they occur. In more advanced designs, such changes can also be used to trigger new assessments or interventions in real time, enabling a shift from static measurement toward dynamic, context-aware research and intervention systems (e.g., Just-in-time Adaptive Interventions).
Passive sensing is part of the broader family of ambulatory assessment methods, which aim to capture real-life data outside the lab room. What distinguishes passive sensing within this framework is its reliance on device-based measurements rather than self-report, combining embedded sensors to approximate patterns of activity, mobility, social context, and physiology. In contrast to Ecological Momentary Assessment (EMA), Experience Sampling Methodology (ESM) or daily diary designs, it therefore does not require active participant input. Instead, it complements these approaches by capturing behavioral and contextual processes that may be difficult, impractical, or impossible to assess through self-report alone.
A helpful way to conceptualize passive sensing is as a continuous data layer underlying everyday life. Rather than sampling discrete moments through repeated assessments, passive sensing generates high-resolution streams of data over time. These data can be used not only to reconstruct patterns of behavior and context, but also to detect meaningful changes as they occur. In more advanced designs, such changes can also be used to trigger new assessments or interventions in real time, enabling a shift from static measurement toward dynamic, context-aware research and intervention systems (e.g., Just-in-time Adaptive Interventions).
Core strengths of passive sensing
Passive sensing shares key features with Ecological Momentary Assessment (EMA) and Experience Sampling Methodology (ESM). It captures data in real-world environments with high ecological validity and produces repeated measurements over time, enabling the study of within-person variability and dynamic processes indaily life.
However, passive sensing differs fundamentally in how data are collected and structured. Its key strengths stem from the fact that measurement is automated, continuous, and independent of active participant input.
However, passive sensing differs fundamentally in how data are collected and structured. Its key strengths stem from the fact that measurement is automated, continuous, and independent of active participant input.
Automatic data collection
Passive sensing does not require active participant input. Data are collected directly from smartphones and wearable devices, enabling measurement to continue independently of participant engagement. This reduces reliance on compliance and minimizes missing data due to non-response. It also allows data collection to persist during moments when participants are unable or unwilling to respond to surveys.
Unobtrusive measurement
Because sensing runs in the background, it does not interrupt participants’ daily activities. This minimizes participant burden and reduces the likelihood that the measurement itself alters behavior, allowing processes to unfold more naturally. As a result, it is particularly suited for studying behaviors that may be disrupted by active assessment (e.g., driving, playing sports, etc.).
Continuous time series
Passive sensing can generate continuous streams of data rather than discrete observations. Because data may be captured continuously, short-lived changes are less likely to be missed, enabling higher temporal resolution and the detection of fine-grained, short-term patterns. At the same time, continuous coverage across days or weeks allows researchers to capture broader behavioral trends and longer-term regularities, resulting in a more complete view of how behavior and context evolve over time.
Long-term and scalable data collection
Because passive sensing requires minimal effort from participants, it is particularly well suited for longer study durations and long-term monitoring. Data can be collected over extended periods, ranging from weeks to months or even years, enabling the study of slow-changing processes, long-term trends, and individual behavioral signatures.
Potential limitations of passive sensing
While passive sensing offers clear advantages in terms of continuous and unobtrusive measurement, it also has some important downsides. Compared to self-report methods like Ecological Momentary Assessment (EMA) and Experience Sampling Methodology (ESM), passive sensing introduces conceptual, technical, and ethical challenges that researchers need to consider carefully.
No direct link between sensor data and psychological processes
Passive sensing captures physiological signals and behavioral patterns, but these do not map directly onto psychological states. This mind-body problem works in two ways.
First, a single sensing pattern can have multiple meanings: for example, reduced mobility may reflect fatigue, focused work, or social withdrawal, depending on the situation. Second, the same psychological state can show up differently in sensing data across people or contexts: for example, stress may lead to increased phone use and restlessness in one person, but to withdrawal, reduced activity, or longer periods at home in another.
As a result, sensor data do not provide direct access to psychological states such as thoughts or emotions, and interpretation often requires combining multiple objective signals and integrating subjective self-report to draw reliable conclusions.
First, a single sensing pattern can have multiple meanings: for example, reduced mobility may reflect fatigue, focused work, or social withdrawal, depending on the situation. Second, the same psychological state can show up differently in sensing data across people or contexts: for example, stress may lead to increased phone use and restlessness in one person, but to withdrawal, reduced activity, or longer periods at home in another.
As a result, sensor data do not provide direct access to psychological states such as thoughts or emotions, and interpretation often requires combining multiple objective signals and integrating subjective self-report to draw reliable conclusions.
Device heterogeneity and battery consumption
Data coverage and quality in passive sensing is affected at multiple levels. First, differences between devices matter: smartphones and wearables vary in type, quality, and availability of sensors, making data from different devices not always directly comparable. For example, activity or physiological measures can differ depending on the hardware and preprocessing algorithms used, introducing unwanted variability across participants.
Second, even within the same device, data collection can be disrupted by operating system restrictions, app termination, and battery optimization processes that limit background activity. These constraints are particularly relevant for smartphones, where sensing competes with other apps and battery use. Wearables are often designed for continuous sensing, but their data quality depends on consistent and correct use.
Second, even within the same device, data collection can be disrupted by operating system restrictions, app termination, and battery optimization processes that limit background activity. These constraints are particularly relevant for smartphones, where sensing competes with other apps and battery use. Wearables are often designed for continuous sensing, but their data quality depends on consistent and correct use.
Privacy and ethical concerns
Passive sensing involves the continuous collection of potentially sensitive data, such as location traces, app usage, and ambient sound. This raises important ethical challenges related to informed consent (i.e., participants clearly understand what data are collected, how they are used, and for what purpose), data minimization (i.e., only collecting the data that are strictly necessary), and secure storage (i.e., protecting data through encryption and controlled access).
More fundamentally, it changes the nature of measurement: instead of relying on what participants actively choose to report, data are passively observed in the background. This shift requires careful attention to transparency, participant control, and ongoing awareness of what is being collected.
More fundamentally, it changes the nature of measurement: instead of relying on what participants actively choose to report, data are passively observed in the background. This shift requires careful attention to transparency, participant control, and ongoing awareness of what is being collected.
Analytical complexity
Passive sensing generates large volumes of high-frequency data that require substantial preprocessing before they can be analyzed. Raw sensor streams need to be cleaned, transformed, and summarized into meaningful features and across different time scales, and there is often no standardized way to do so. As a result, analytical choices can strongly influence outcomes, increasing the complexity of analysis and raising challenges for transparency, reproducibility, and comparability across studies.
Common sensors and approaches in passive sensing
Passive sensing typically relies on two types of devices: smartphones and wearables (e.g., smartwatches or -rings). While both collect real-world data, they differ in what they measure and how they are used. In practice, all sensors provide a specific type of signal, which can be translated into meaningful indicators of real-world functioning. Furthermore, both smartphone and wearable sensing can be implemented at different levels of intensity.
Smartphones versus wearables
Smartphones are particularly strong in measuring context and behavior:
- GPS: location, mobility patterns, time spent at home versus outside
- Accelerometer: movement, activity levels, sedentary behavior
- Screen unlocks and app usage: phone engagement (calls, messages, app interactions)
- Bluetooth: proximity to others, social context proxies
- Microphone: environmental context such as noise levels
- Wi-Fi: public versus private space
Wearables are well-suited for physiological and activity-related signals:
- Heart rate: physiological arousal, stress-related patterns
- Heart rate variability (HRV): recovery and regulation processes
- Step count and activity intensity: physical activity and energy expenditure
- Sleep metrics: duration, timing, and fragmentation of sleep
Full versus lite sensing
Full sensing is defined as continuous background data collection across multiple sensors, providing high-resolution, real-time streams of behavior, context, and physiology. This offers a very complete picture of daily life, but may feel more invasive, impact battery life, and require more complex setup due to system restrictions.
In contrast, lite sensing means that the sensor data are only collected at specific moments, such as during EMA / ESM assessments. This reduces battery usage and is typically less intrusive, while giving more control over when data are captured. However, it provides only momentary snapshots or sensing aggregates between assessments.
The choice between full and lite sensing depends on the research question, privacy considerations, and practical constraints, reflecting a trade-off between data richness and feasibility.
In contrast, lite sensing means that the sensor data are only collected at specific moments, such as during EMA / ESM assessments. This reduces battery usage and is typically less intrusive, while giving more control over when data are captured. However, it provides only momentary snapshots or sensing aggregates between assessments.
The choice between full and lite sensing depends on the research question, privacy considerations, and practical constraints, reflecting a trade-off between data richness and feasibility.
The relation between passive sensing and EMA / ESM
As previously discussed, sensing and Ecological Momentary Assessment (EMA) / Experience Sampling Methodology (ESM) share real-world, repeated measurement but differ in how data are collected: passive sensing versus active self-report. This raises the question of how these two data streams may relate in practice.
First, passive sensing and EMA / ESM can be treated as complementary sources of information. Sensing captures objective indicators of behavior and context, while EMA / ESM captures subjective experiences such as thoughts, emotions, and appraisals. Combining both allows researchers to link what people do and where they are to what they feel, leading to a more complete understanding of real-life processes.
Second, they can be partially interchangeable. In some cases, sensor data can replace or reduce certain self-report questions, such as those related to activity (e.g., What are you doing right now?), location (e.g., Where are you right now?), or device use (e.g., How many minutes have you used your smartphone since the last prompt?), thereby lowering participant burden. However, this interchangeability is limited to constructs that can be reasonably approximated by sensors and does not extend to subjective experience.
Finally, passive sensing can be used to support and inform EMA / ESM study design. It can help define more efficient sampling schemes by triggering assessments at meaningful moments, such as changes in behavior or context. This shifts measurement from traditional time-based schedules toward context-aware designs that focus on when something relevant actually happens.
First, passive sensing and EMA / ESM can be treated as complementary sources of information. Sensing captures objective indicators of behavior and context, while EMA / ESM captures subjective experiences such as thoughts, emotions, and appraisals. Combining both allows researchers to link what people do and where they are to what they feel, leading to a more complete understanding of real-life processes.
Second, they can be partially interchangeable. In some cases, sensor data can replace or reduce certain self-report questions, such as those related to activity (e.g., What are you doing right now?), location (e.g., Where are you right now?), or device use (e.g., How many minutes have you used your smartphone since the last prompt?), thereby lowering participant burden. However, this interchangeability is limited to constructs that can be reasonably approximated by sensors and does not extend to subjective experience.
Finally, passive sensing can be used to support and inform EMA / ESM study design. It can help define more efficient sampling schemes by triggering assessments at meaningful moments, such as changes in behavior or context. This shifts measurement from traditional time-based schedules toward context-aware designs that focus on when something relevant actually happens.
Bridging the mind-body gap in passive sensing
A central challenge in passive sensing is translating raw sensor data into meaningful indicators of psychological processes. Because sensor data are not directly interpretable, this requires a series of transformation steps, each involving assumptions and design choices. This can be illustrated with a running example: using smartphone data to understand changes in depressive mood.
From raw signals to features
Sensor data are first collected as raw streams (e.g., GPS coordinates, accelerometer values, phone usage logs). These raw data are then processed into features that summarize behavior over a given time window. For example, GPS data can be transformed into time spent at home and distance traveled, accelerometer data into activity levels, and phone logs into total screen time or number of unlocks.
From features to behavioral indicators
Next, these features are translated into more interpretable behavioral indicators. In this example, GPS and activity data can be used to estimate mobility and daily structure, while phone usage patterns can reflect engagement or inactivity. Together, these indicators begin to describe how someone functions in daily life.
From behavioral indicators to digital phenotypes
In a further step, multiple behavioral indicators are combined into higher-level patterns, often referred to as digital phenotypes. For instance, a combination of reduced mobility, increased time at home, lower activity levels, and increased or irregular phone use may form a digital phenotype indicative of withdrawal or disrupted daily routines.
Linking digital phenotypes to psychological constructs
Finally, these digital phenotypes are linked to psychological variables such as mood. For example, this pattern of reduced mobility and altered phone use may be associated with lower mood or emerging depressive symptoms, especially when validated against self-report data. Importantly, this relationship is indirect and probabilistic rather than deterministic.
Because each step involves assumptions, validation is essential. Researchers should test whether mobility-based digital phenotypes relate to reported mood and refine their models accordingly.
Because each step involves assumptions, validation is essential. Researchers should test whether mobility-based digital phenotypes relate to reported mood and refine their models accordingly.
Roots and history of passive sensing
Passive sensing did not emerge as a standalone method, but developed at the intersection of several scientific and technological trends within the last two decades.
From ambulatory assessment to digital sensing
Passive sensing builds on the tradition of ambulatory assessment, which aims to capture real-life processes outside the laboratory through repeated self-report methods such as diary studies and later Ecological Momentary Assessment (EMA) or Experience Sampling Methodology (ESM). The widespread adoption of smartphones and wearable devices extended this idea by enabling continuous, device-based measurement of behavior, context, and physiology in daily life, shifting from active reporting to passive data collection.
The "quantified self" and precision medicine
At the same time, the quantified self movement, founded by Gary Wolf and Kevin Kelly in 2007, promoted the idea that individuals can gain insight into their lives through systematic self-tracking. By collecting data on activity, sleep, physiology, and behavior, this movement emphasized personal data as a tool for self-knowledge and behavioral change.
More recently, passive sensing has become closely linked to developments in precision or personalized medicine, which aim to understand and predict individual-specific patterns rather than relying on group-level averages. By continuously capturing data in daily life, mobile sensing supports more personalized, data-driven approaches to research and care.
More recently, passive sensing has become closely linked to developments in precision or personalized medicine, which aim to understand and predict individual-specific patterns rather than relying on group-level averages. By continuously capturing data in daily life, mobile sensing supports more personalized, data-driven approaches to research and care.
Dr. Egon Dejonckheere
Egon Dejonckheere is an Assistant Professor at Tilburg University, affiliated with KU Leuven, specializing in ambulatory assessment and intensive longitudinal research. His work combines methodological innovation to improve these methods with applications in mental health and emotion dynamics, and includes dozens of smartphone-based studies in daily life.
FAQ
01.
What is passive sensing?
Passive sensing is a research method that uses smartphones and wearable devices to automatically collect data on behavior, context, and physiology in daily life. Unlike surveys, it does not require active input from participants and can capture data continuously in real-world environments.
02.
How is passive sensing different from EMA or ESM?
Passive sensing enables continuous, real-world data collection with minimal participant burden. It provides high-resolution time series data, captures behavior that is difficult to self-report autonomously and without interference, and allows for long-term monitoring of daily life patterns.
03.
What are the limitations of passive sensing?
Key limitations of passive sensing include the lack of a direct link between sensor data and psychological states, variability across devices, battery and technical constraints, privacy concerns, and the complexity of analyzing large, high-frequency datasets.
04.
What types of data can passive sensing collect?
Passive sensing can collect data such as location (GPS), movement (accelerometer and gyroscope), smartphone usage (app interactions, keyboard strokes), social proximity (Bluetooth), heart rate (variability), activity levels, and sleep patterns. The exact data depend on the sensors available on smartphones and wearable devices.
05.
What is a digital phenotype?
A digital phenotype is a pattern of behavior derived from sensor data. It often combines multiple indicators, such as mobility, activity, and phone use, to approximate psychological processes like mood, stress, or symptoms in daily life.
06.
Can passive sensing measure psychological states directly?
No, passive sensing does not directly measure thoughts or emotions. Instead, it provides indirect indicators based on behavior and physiology. These indicators need to be interpreted carefully and are often combined with self-report data to draw reliable conclusions.
07.
What is precision medicine?
Precision or personalized medicineis a healthcare and research approach that tailors prevention and treatment to the individual rather than relying on group averages (i.e., one-size-fits-all approach). By using detailed, often continuous data, such as lifestyle data and sensor-based measurements, it seeks to identify individual-specific patterns and deliver more targeted and effective care.