Wearable sensors are providing exciting new research opportunities for the social sciences. Following up on initial technical developments to miniaturize and combine several sensor technologies into wearable devices in the first decade of the 21^st century, interested scholars have invested considerable effort in assessing the validity and reliability of the resulting data (Chaffin et al., 2017; Chen & Miller, 2017; Elmer et al., 2019; Kayhan et al., 2018). These initial studies relied mainly on laboratory experiments to assess the validity of sensors as indicators of physical constructs such as ‘proximity’ based on Bluetooth (BT) signals or ‘face-to-face’ detections based on infrared sensors. However, an increasing number of studies that deploy wearable sensors in real-world organizational settings have become available. This allows the variability of the sensor measurements to be assessed under realistic settings beyond controlled laboratory environments, while also putting the focus on the suitability of sensor data as indicators of higher-level social and psychological constructs. Several available studies have explored sensor data as indicators of “creativity” (Parker et al., 2018), “friendship” and “advice-seeking” (Matusik et al., 2018), or “subjective wellbeing” (Alshamsi et al., 2016) and “happiness” (Yano et al., 2015).

While studies based on real-world field settings make important contributions to assess the variability of sensor data and higher-level constructs, the existing variety of empirical field settings has been rather meager to date. Usually, wearable sensors are deployed in a single, relatively large group of people working together. The resulting findings are thus limited to one specific group and field situation without any means of extrapolating to other groups and/or conditions. This article addresses this problem by analyzing and comparing wearable sensor data among nine relatively small research and development (R&D) teams.

As a result, we therefore firstly offer important insights into the inter-group variability of Sociometric^Footnote1 measurements for relatively similar R&D teams. By examining in more detail how important metrics vary between the nine comparable R&D groups, a more finely tuned picture of the context-sensitive nature of supposedly ‘objective’ sensor measurements begins to emerge. Secondly, our research also contributes to the important task of validating Sociometric, sensor-based measurements for higher-level constructs. BT signals are usually taken as a measurement of physical proximity between devices (or the people wearing them), which in turn should ideally provide a valid indicator of social ties such as friendship. Whereas others have shown that there is a moderate relationship between BT signals and these social ties, this article provides further insights into the strength of the relationship between Sociometric proximity and friendship on the one hand, and proximity and advice-seeking on the other, taking into account the different organizational embedding of groups. With regard to BT values, we can ask how reliably certain radio signal strength indicator thresholds discriminate between friendship or advice networks, not only within the same group but across several groups. By inserting important organizational and team-based control variables, we show how these thresholds might vary according to the wider context.

The article is structured as follows: in the first section we will briefly summarize the main findings of existing research using wearable sensors. This includes both laboratory validation studies as well as field research focusing on higher-level constructs. To the best of our knowledge, there has thus far been no study that analyzes sensor data across several comparable groups. Next, in the Methods section, the details of the field research are described in conjunction with the important data pre-processing steps carried out. In addition, there are also some initial sketches of the socio-demographic and sociometric profiles of the participating teams and an introduction to the overall analytical approach. The third section then describes the overall results, followed by a general discussion of their implications before we conclude the article with some final remarks and recommendations.

To read the full article prepared by Jörg Müller, Julio Meneses, Anne Laure Humbert & Elisabeth Anna Guenther follow this link below to the Springer website

https://link.springer.com/article/10.3758/s13428-020-01444-x