An experimental laboratory study was conducted in which children performed a variety of digital technology–related and non–digital technology–related tasks while wearing a wearable camera and being video recorded by 2 room video cameras within a 10-m-diameter laboratory space.

Typically developing children aged between 3 and 14 years were recruited among university networks and in the community via word of mouth, flyers, and social media advertisements. Recruitment was stratified based on gender and age of the participants, with the aim of recruiting approximately equal numbers of boys and girls and children in 3 age brackets according to Australian school ages (3‐5 y to represent a preschool-aged population, 6‐10 y to represent a primary school–aged population, and 11‐14 y to represent secondary school–aged and early adolescence population). Children were excluded from the study if they had any psychological or physical clinical diagnosis, as reported by their caregiver, which may have influenced their ability to understand and follow instructions or perform requested tasks.

Ethical approval was obtained from the Curtin University Human Ethics Committee (HRE2022-0157). Caregivers provided written informed consent, and children gave verbal or written assent, as appropriate to their age.

Children attended a single approximately 1-hour data collection session at a Curtin University laboratory with their caregivers between July and October 2022. Children either participated in the data collection session alone (n=10) or with a friend or sibling (n=38), with a maximum of 2 children participating simultaneously. Two to three researchers, experienced in working with children, were present at each data collection session. After providing informed consent, caregivers completed a brief sociodemographic questionnaire. Subsequently, children were asked for their verbal or written assent, as appropriate to their age, and were then fitted with the wearable camera.

The laboratory was set up to capture different tasks, both with and without digital technology, and involving a range of postures and movements. Following laboratory and equipment familiarization, children then completed each task for at least 2 to 3 minutes, allowing for 120 to 180 images to be captured by the wearable camera (based on a 1-s epoch frequency). Tasks were performed in a standardized order where possible. However, flexibility was given in the protocol to ensure the participant’s enjoyment. Some tasks were completed with a research assistant, the child’s caregiver, or the second child in the session to replicate coviewing tasks (eg, when a child watches a television program with a caregiver or sibling) and to support children in completing the tasks. Tasks included watching programs on a television or laptop, playing on a gaming device (Nintendo Switch), reading a book, drawing on paper, playing with magnetic tiles or Jenga, using a desktop computer, drawing on a tablet, taking photographs with a smartphone, throwing and catching a ball, watching a video on a smartphone, and checking step counts via a smartwatch. Children were instructed on the position or movement to complete the task to allow the capture of images and video from a range of ecologically valid postures and movements. Table 1 provides a full list of tasks completed by the child, with devices used for each task. In total, 9 technologies were included: television, desktop, laptop, tablet, smartphone, smartwatch, handheld gaming console, combined gaming controller, and single gaming controller.

A total of 48 children completed the study; however, a technical error resulted in some missing wearable camera image time stamps. Therefore, 4 participants were excluded, leaving a sample of 44 participants (girls: n=25, 55%) with a mean age of 8.3 (SD 3.3) years (Multimedia Appendix 2). Eleven children were aged between 3 and 6 years, 22 children were aged between 7 and 10 years, and 11 children were aged between 11 and 14 years. Across the 44 participants, a total of 182,451 contemporaneous wearable camera images and room video frames were coded, which equated to 50.7 hours of observed recording. At an individual participant level, the mean recording duration was 72.5 (SD 11.2, range 39.7-103.0) minutes.

Table 3 presents the frequency distribution of categories with at least 100 images or frames using the 53 categories of possible technology codes based on the wearable camera images and room video frames. Table S2 in Multimedia Appendix 1 contains the distribution of all 53 categories. “No digital technology” was coded for 46% (n=84,178) and 29% (n=52,504) of the wearable camera images and room video frames, respectively. The most frequently coded technologies based on the wearable camera images were “television” (n=21,900, 12%), “laptop” (n=11,596, 6%), and “desktop” (n=9426, 5%). For the room video frames, the most frequently coded technologies were “laptop” (n=19,150, 11%), “television” (n=18,266, 10%), and “television & single gaming controller” (n=16,005, 9%). Ten percent (n=17,802) of the wearable camera images were coded as “uncodable” compared to only 3 (0%) room video frames.

Table 4 presents the frequency distribution of categories using the 9 “any technology combination” category codes for both the wearable camera images and room video frames. The most frequently coded technologies based on the wearable camera images were “any television” (36,198/182,451, 19.8%), “any laptop” (n=18,816, 10.3%), and “any desktop” (n=9479, 5.2%). For the room video frames, the most frequently coded technologies were “any television” (n=61,477, 33.7%), “any laptop” (n=29,461, 16.1%), and “any tablet” (n=18,075, 9.9%).

There were similar frequencies of codes between genders and age groups for the 9 “any technology combination” coded based on the wearable camera and room video (Table S3 in Multimedia Appendix 1). Overall, there were no significant differences in the PABAK results between either gender or age groups (Table S3 in Multimedia Appendix 1).

Overall, there were generally very low levels of agreement between the codes based on wearable camera images and room video frames for all technology types, as well as for coding whether or not technology was being used. Across all the categories, the highest agreement was found for “desktop” and “handheld gaming console” codes, and the lowest agreement was found for “smartwatch.” Differences mostly occurred when multiple technologies were being used or coded or not, as combinations of technology were often captured with the room video but not necessarily with the wearable camera. Similar patterns to those observed in the whole sample were seen when comparing genders and age groups.

Previous literature has compared self-reported and objective measurements of technology use [35,36], with a previous systematic review finding discrepancies between device-logged and self-reported digital media use and reporting that self-reported measures were rarely accurate reflections of device-logged media use [37]. While there are now many studies that have used visual images to measure human behavior, no previous studies that have directly compared wearable camera images and room video frames to understand what can be captured from each of these data sources were identified. The wearable camera and room video varied in aspects such as sampling rate, the ability of the camera to move, and field of view, leading to differences in what was captured by each source. Prior studies have compared different types of cameras and camera views for other outcomes, for example, comparing a head-worn camera with a laptop camera to assess eye tracking [38] and a head-worn camera with an overhead drone camera to assess park quality [39]. The closest identified comparison to this study compared measures of parent-infant behavior using the same head-worn video cameras on the infant, parent, and observing researcher [40]. This study found that both the direction of the field of view (eg, infant’s perspective and mother’s perspective) and the viewing distance (ie, researcher perspective vs infant and mother) influenced what could be measured. Their findings thus concur with this study that while there was concordance on some outcomes, different camera views provided better options for different outcomes. For example, infant facial expressions were best measured from the mother-worn camera; however, due to the differing visual field, whole-body movements were often missed by the first-person cameras (mother- and infant-worn camera) compared to the third-person cameras (researcher perspective) [40].

The highest agreement for technology types was observed for technology devices that were used when the child was more stationary or when the device was closer to the child (ie, desktop and handheld gaming consoles; Figure 2). Brief technology activities (eg, smartwatch use) were captured better with the room video due to higher sampling rate, resulting in the lowest agreement (Figure 3). Child engagement with a smartwatch tended to occur quickly and was therefore easily missed with the 1-second sampling frequency of camera images and may also have been missed in camera images because of blurring due to the quick movement. Smartwatches are becoming increasingly available to children; therefore, it would be important to capture use effectively [41]. In addition, when a child looked at their wrist, it may not be in view of the wearable camera (eg, the child’s arm blocking the camera). Due to these multiple issues, the use of the smartwatch may not have been captured in the wearable camera, which could perhaps be better captured in a room video recording at 30 frames per second.

Activities involving child movement (eg, jumping, running, skipping, and cartwheels) were not captured as well by the wearable camera compared to the room video. A higher percentage of images were coded as “uncodable” for the wearable camera images compared to the room video frames. This was often due to movement of the child or camera and a slow shutter speed, leading to blurred images. Furthermore, activities involving prone postures were not captured well by the wearable camera, as the child blocked the wearable camera view (ie, child lying on their stomach; Figure 4).

Owing to the differing fields of view, the wearable camera could miss technology but was more likely to detect active use of technology. It is likely that background use of certain technologies (ie, television and laptop) was coded based on the room video recording but not by the wearable camera images. More frames were coded as “no digital technology” based on the wearable camera images compared to the room video recording. Either the wearable camera missed capturing the technology or the room video “overcaptured” technology use. As captured by the room video, technology devices could still have been used, but due to the field of view of the wearable camera, the technology was missed. However, when determining active technology use based on the room video recording, it was sometimes hard to determine if the child was actively using or engaged with the technology or not, for example, if a technology was in the peripheral vision of the child (Figure 5). Wearable cameras primarily capture first-person, action-proximal engagement, whereas room video captures environmental availability and background exposure.

Technology use involving simultaneous use of hand gaming controllers and a television screen was not captured as well by the wearable cameras due to the field of view missing either the controller or television screen. Single and combined gaming controllers were often coded simultaneously with television for the room video frames but often only coded as “television” based on the wearable camera images (Figure 6). If single and combined gaming controllers were simultaneously coded with television in the wearable camera, then they were likely also simultaneously coded in the room video. The single and combined gaming controllers were often held outside the field of view of the wearable camera and, when within the field of view, could appear blurry in the wearable camera images due to movement of the child and controllers and a slow shutter speed. On the basis of the wearable camera, it was sometimes difficult to determine whether the child was simply watching television or gaming using the television, whereas this could be captured within the room video given the different point of view.

Room video data had a wider field of view; therefore, it was able to better capture technology that was in the room and had the ability to capture the context of the technology use. With the ability to capture the whole room, it was clear when the child was passively watching a program on the television compared to actively playing a video game. Additionally, although not an aim of this study, room video would have the ability to capture coviewing and co-use of technology (as seen in Figures 2-7). Borzekowski and Robinson [16] previously used a fixed room camera to capture coviewing of television and behaviors while watching television, such as eating, reading, or doing homework, and playing. On the other hand, the wider field of view of the video recording sometimes limited the quality of the video frames, and therefore, on occasion, it was hard to determine if there was technology present (eg, a laptop in the background). In addition, although not an aim of this study, it would be difficult to determine the content of the technology. For example, on occasion, it was difficult to determine if the television was actively on with a program or just a standby screen; therefore, it would be difficult to determine if the content was a violent movie compared to educational children’s content. Furthermore, when determining active technology use based on the room video, it was sometimes difficult to determine if the child was actively using or engaged with the technology or not, for example, capturing technology in the peripheral vision of the child (as seen in Figure 5). Vadathya et al [21] previously attempted to combat this issue by developing a method to determine whether the target child was looking at the television using a home-based fixed room camera. There are also privacy concerns with recording room images.

The wearable camera had a smaller field of view but was closer in proximity and more focused (as seen in Figures 2 and 7). The wearable camera was able to capture active engagement with the technology and, due to the proximity to the technology, would have a greater ability to capture the content on the screen (as seen in Figure 7). Previous studies using wearable cameras to measure device use by adolescents have shown that it is feasible to code multiple screens with details of the screen content (eg, gaming, television viewing, social media, communication, creative content, and internet) as well as the location (eg, bedroom and living room) [18,19]. However, a wearable camera also has limitations. The children were often moving, which led to blurry images (as seen in Figure 4). With the camera mounted on the chest, sometimes some technologies were not visible either due to the position of the child or the placement of the technology (as seen in Figure 6). Particularly for gaming, the controller was frequently not visible in the wearable camera frame due to the child holding the controller out of view or due to movement of the child (such as using the single controller during active play). Therefore, the task may have been missed, with both gaming and television watching activities being coded as “television.” Sometimes only a very small part of the screen was visible (eg, the corner of a tablet), therefore potentially missing technology that was out of frame. Moreover, if the engagement with the technology occurred quickly (eg, using a smartwatch), it was easily missed with images only captured every second (as seen in Figure 3). Furthermore, if the child was in certain positions (eg, lying on their stomach), the camera could be blocked, leading to “uncodable” images (as seen in Figure 4). Previous studies have also found that wearable cameras have the ability to capture screen use when the participant is upright or partially upright but capture a high percentage of images of the ceiling or obscure images due to the participants being positioned on their backs or in other positions [18,42,43]. Despite an ethical framework being established for privacy concerns with wearable cameras [28], parents remained concerned, especially with the varied locations of use creating greater potential for capture of nonconsenting third parties. The advantages and challenges of each data source are summarized in Table 5.

A strength of this study was the simultaneous use of both a wearable camera and 2 fixed room video cameras to capture technology use. Another strength was that there was the broad age range from 3 to 14 years and the wide variety of technology devices included in the study. However, there were several limitations. First, the video camera placement at approximately 6 m from the center of the laboratory led to reduced video frame quality. There was also a potential that the time stamp synchronization of the wearable camera images and room video frames was not perfect and that small discrepancies in the time stamp synchronization could result in mismatches, which would affect the agreement of brief activities, such as checking the smartwatch. Finally, as the data were collected in a laboratory setting, the results should be generalized with caution to free-living conditions. We encourage further research in free-living conditions, despite the challenges of such studies.

If future research is continued in free-living conditions, a few considerations should be made. Similar to previous literature [28], caregivers expressed concerns about privacy if the devices were used in a home setting and were worried about other children being captured if the devices were used in a school setting. Caregivers were more comfortable if the recording devices could be turned on and off and if the devices could be smaller. There were also concerns from caregivers about wearing the camera during certain activities, such as contact sports and swimming.

The coding of the wearable camera and fixed video frames was based on decisions made by the coders. In practice, researchers are likely to be able to interpolate coding for some blurred or obscured images, for example, when a screen temporarily moves out of view. This approach was not used in this study to enable clear comparisons of issues, such as movement blurring. Coding of both sources had the potential for human error and was extremely time-consuming (taking ~6 h per hour of recorded data), which suggests that it may be unsuitable for large-scale studies. Machine learning software could potentially be used to reduce the high researcher burden for the analysis of technology use via camera images or video frames [7], and advances in AI should be examined for accuracy in coding child behaviors from images [44].

Given the complexity of modern technology use, future research could usefully evaluate the simultaneous use of several data collection methods to use the relative advantages of different methods [7]. This could include questionnaires to provide some complementary information for image data, such as context or purpose. Images from both the wearable camera and the stationary video camera can capture useful and complementary information and thus could be used together. Similarly, multiple objective measures could be used simultaneously, for example, combining on-device recording of apps used or websites visited to triangulate the classification of content coded from image data.

Overall, there was poor agreement for children’s technology use between the wearable camera images and room video frames. The results did not differ based on gender or age of the participants. The wearable camera and room video varied in sampling rate, the ability to move, and perspective, leading to differences in what was captured by each method. Both the room video and wearable camera methods have strengths and limitations due to these differences. Overall, fixed room video might better capture that technology is within a single set space and the context of the technology use, and the view and higher sampling rate of video may be better when child movement or prone lying is involved, handheld devices are used, interactions with technology are brief, or multiple devices are used within a set location. Wearable cameras could better differentiate the active use of technology versus background television, the content of the technology (due to the closer view of screen detail), and different locations of use. Future use of wearable cameras and room video to capture digital technology use should carefully consider the purpose of the study along with the potential limitations of each data source. Capturing digital technology use by children is difficult, with no gold standard and no one ideal method.