Bluetooth signal strength (RSSI) as a function of distance.
A: Scans between two phones. Measurements are per distance performed every five minutes over the course of 7 days. Mean value and standard deviation per distance are respectively    and . B: Average of the values in respective time-bins. Summary statistics are:    and . C: Maximal value per time-bin. The mean value and standard deviation per distance are:  , , and  The measurements cover hypothetical situations where individuals are far from each other and on either side of a wall.

Bluetooth signal strength (RSSI) as a function of distance. A: Scans between two phones. Measurements are per distance performed every five minutes over the course of 7 days. Mean value and standard deviation per distance are respectively and . B: Average of the values in respective time-bins. Summary statistics are: and . C: Maximal value per time-bin. The mean value and standard deviation per distance are: , , and The measurements cover hypothetical situations where individuals are far from each other and on either side of a wall.

Source publication
Article
Full-text available
Understanding how people interact and socialize is important in many contexts from disease control to urban planning. Datasets that capture this specific aspect of human life have increased in size and availability over the last few years. We have yet to understand, however, to what extent such electronic datasets may serve as a valid proxy for rea...

Citations

... Likewise, possessing long-and short-range networks recognizable permits us to avoid formation of man-made networks by randomization systems. The formation of network of long-range is via connections happening at any space enclosed via range of Bluetooth, amid 0.0 and 10.0 to 15.0 meters, While the creation of short-range network via choosing the subset of connections with ≥-75dbm equivalent to spaces of roughly 1.0 meter or fewer with the purpose of capture only near range connections as displayed in Figure 1 [21]. The link weights (i, j) are broadly distributed Examination via reviewing variances and likenesses in the link weights spreading amid the three networks (short-range, sampled long-range, and long-range). ...
... The weight of a link linking two persons is definite as the whole number of connections happening on that link ∑ . It is probable to compute the spreading of weight for this network methodically, as the tested long-range network is produced via sampling connections at arbitrary from the complete network [21], make significant variances amid the networks of long -range and the short -range and can calculate this variance utilizing the Shannon entropy. For a node , begin from a connection with neighbor j with weight and express ( ) ∑ ⁄ to show the fraction of the connections of whole of node happening on that link. ...
Article
Full-text available
The getting of knowledge from the raw data has delivered beneficial information in several domains, the prevalent utilizing of social media produced extraordinary quantities of social information. Simply, social media delivers an available podium for employers for access to information. Data Mining has ability to present applicable designs that can be useful for employers, commercial, and customers. Data of social media are strident, massive, formless, and dynamic in the natural case, so modern encounters grow. Investigation methods of data mining utilized via social networks is the purpose of the study, accepting investigation plans on the basis of criteria, and by selecting a number of papers to serve as the foundation for this article. Afterward a watchful evaluation of these papers, it has been discovered that numerous data extraction approaches were utilized with social media data to report a number of various research goals in several fields of industrial and service. Though, implementations of data mining are still raw and require more work via industry and academic world to prepare the work sufficiently. Bring this analysis to a close. Data mining is the most important rule for uncovering hidden data in large datasets, especially in social network analysis, and it demonstrates the most important social media technology.
... To corroborate our theory, we performed numerical simulations upon a temporal contact network estimated via Bluetooth signal exchanges in the Copenhagen Networks Study [40]. We retained those interactions with an associated Received Signal Strength Indication (RSSI) not lower than −74 dBm, corresponding to physical distances approximately up to 2 meters [41]. The resulting temporal network involves 672 individuals and 374884 pairwise interactions spread over 8064 timestamps, binning four weeks of recording time into five-minute intervals. ...
Preprint
Full-text available
Physical contacts do not occur randomly, rather, individuals with similar socio-demographic and behavioural characteristics are more likely to interact among them, a phenomenon known as homophily. Concurrently, the same characteristics correlate with the adoption of prophylactic tools. As a result, the latter do not unfold homogeneously in a population, affecting their ability to control the spread of infectious diseases. Here, focusing on the case of vaccines, we reveal three different dynamical regimes as a function of the mixing rate between vaccinated and non vaccinated individuals. Specifically, depending on the epidemic pressure, vaccine coverage and efficacy, we find the attack rate to decrease, increase or vary non monotonously with respect to the mixing rate. We corroborate the phenomenology through Monte Carlo simulations on a temporal physical contact network. Besides vaccines, our findings hold for a wide range of prophylactic tools, indicating a universal mechanism in spreading dynamics
... Sensor data, such as heart rate, sleep patterns and smartphone-based proximity measures may be of little relevance on their own, but they can be important measures when combined with other data. Sekara and Lehmann (2014) used, for instance, proximity sensor data to study the strength of friendship ties. ...
Chapter
Full-text available
Digital data and methods are becoming increasingly ubiquitous. Almost all aspects of people's lives are now digitized: population registers are digital; health records are digital, criminal records are digital, employment and education registers are digital, all our interactions with authorities (whether on a local, regional or national level) are increasingly digital and hence produce digital records; most of our economic transactions (purchases, bank transfers, sales) are increasingly digital and hence produce digital records; our entertainment consumption is increasingly digital (Netflix, Spotify) and hence produce digital records, and even our social life becomes increasingly digitized with social media like Facebook, Instagram, WhatsApp etc. While there are still important realms that remain off the digital grid (e.g. elections in most countries, real-life social interactions etc.), the vastness of activities that are now digitized can hardly be ignored. This creates both opportunities and challenges for social science researchers and sociologists in particular (Golder and Macy 2014, Hampton 2017). In this chapter I will discuss how analytical sociology can help to harness the opportunities and respond to the challenges and why and how analytical sociology should embrace digital data and methods in its repertoire of methodological approaches. I will first define and discuss what digital data and digital methods are, then I will discuss the opportunities and challenges and how analytical sociology can help to harness the first and deal with the latter, and finally I will discuss what digital data and digital methods can offer to analytical sociology.
... where time t is in units of days and s ij (t) is the maximum Bluetooth signal strength between participants i and j measured during day t, while measurements where performed every five minutes. The threshold 80 dBm corresponds to a distance of about 2 m and maximises the ratio of social interactions to transient and unimportant connections [76]. To minimise noise from the beginning and end periods of data collection, i.e. noise due to participants joining late or dropping out early, in this study we focus on the period from the first of February 2014 to the end of April 2014, which corresponds to the spring semester and is in the middle of the "SensibleDTU 2013" data collection, the second deployment of CNS. ...
Article
Full-text available
Spreading dynamics and complex contagion processes on networks are important mechanisms underlying the emergence of critical transitions, tipping points and other non-linear phenomena in complex human and natural systems. Increasing amounts of temporal network data are now becoming available to study such spreading processes of behaviours, opinions, ideas, diseases and innovations to test hypotheses regarding their specific properties. To this end, we here present a methodology based on dose–response functions and hypothesis testing using surrogate data models that randomise most aspects of the empirical data while conserving certain structures relevant to contagion, group or homophily dynamics. We demonstrate this methodology for synthetic temporal network data of spreading processes generated by the adaptive voter model. Furthermore, we apply it to empirical temporal network data from the Copenhagen Networks Study. This data set provides a physically-close-contact network between several hundreds of university students participating in the study over the course of 3 months. We study the potential spreading dynamics of the health-related behaviour “regularly going to the fitness studio” on this network. Based on a hierarchy of surrogate data models, we find that our method neither provides significant evidence for an influence of a dose–response-type network spreading process in this data set, nor significant evidence for homophily. The empirical dynamics in exercise behaviour are likely better described by individual features such as the disposition towards the behaviour, and the persistence to maintain it, as well as external influences affecting the whole group, and the non-trivial network structure. The proposed methodology is generic and promising also for applications to other temporal network data sets and traits of interest.
... A high spatiotemporal resolution is necessary to faithfully simulate disease transmission through a social network, since diseases (such as COVID-19) may be less likely to transmit during short encounters or between individuals separated by more than a few meters [32,33]. The upper limit for the range of our Bluetooth data is approximately 15 m [34]. We also note that our data are similar in nature to those collected by contact tracing smartphone applications [35]. ...
... The approximate distance between participants can be inferred from the strength (RSSI) of the Bluetooth signal transmitted between devices. The signal strength can resolve distances in the range of ≤ 1 meter to approximately 10-15 m [34]. To prepare our data for modeling of disease transmission, the collected RSSI values are related to an epidemiologically relevant notion of contact. ...
... Our transmission model assumes that the transmission risk of COVID-19 increases sharply as interpersonal distance is decreased below 1-2 m [33,[40][41][42][43]. Thus, we define two individuals to be in social contact whenever the Bluetooth signal strength between their respective devices exceeds − 85 dBm. This definition of contact captures essentially all ≤ 1m interactions while excluding a large portion of the 3m interactions and above [34]. ...
Article
Full-text available
Digital contact tracing has been suggested as an effective strategy for controlling an epidemic without severely limiting personal mobility. Here, we use smartphone proximity data to explore how social structure affects contact tracing of COVID-19. We model the spread of COVID-19 and find that the effectiveness of contact tracing depends strongly on social network structure and heterogeneous social activity. Contact tracing is shown to be remarkably effective in a workplace environment and the effectiveness depends strongly on the minimum duration of contact required to initiate quarantine. In a realistic social network, we find that forward contact tracing with immediate isolation can reduce an epidemic by more than 70%. In perspective, our findings highlight the necessity of incorporating social heterogeneity into models of mitigation strategies. Graphic abstract
... Similar data collection strategies and the use of Bluetooth proximity for finding real-world group gatherings were used and validated by others in 8,[24][25][26][27][28] . Table 1 lists the details about the groups identified from the Bluetooth data. ...
Article
Full-text available
Understanding why people join, stay, or leave social groups is a central question in the social sciences, including computational social systems, while modeling these processes is a challenge in complex networks. Yet, the current empirical studies rarely focus on group dynamics for lack of data relating opinions to group membership. In the NetSense data, we find hundreds of face-to-face groups whose members make thousands of changes of memberships and opinions. We also observe two trends: opinion homogeneity grows over time, and individuals holding unpopular opinions frequently change groups. These observations and data provide us with the basis on which we model the underlying dynamics of human behavior. We formally define the utility that members gain from ingroup interactions as a function of the levels of homophily of opinions of group members with opinions of a given individual in this group. We demonstrate that so-defined utility applied to our empirical data increases after each observed change. We then introduce an analytical model and show that it accurately recreates the trends observed in the NetSense data.
... For our study we defined an encounter as physical proximity as measured by a smartphone via a Bluetooth measurement. We used a Bluetooth signal of À80 dBm or stronger to indicate encounters as Sekara and Lehmann (2014) showed this to be a reliable cut-off value for close and unobstructed physical proximity for this dataset. Given that we were only interested in time spent at stop locations, this meant an encounter in our study represented either the physical co-location of two students in the same room or in close proximity outdoors. ...
Article
Full-text available
Space, time and the social realm are intrinsically linked. While an array of studies have tried to untangle these factors and their influence on human behaviour, hardly any have taken their effects into account at the same time. To disentangle these factors, we try to predict future encounters between students and assess how important social, spatial and temporal features are for prediction. We phrase our problem of predicting future encounters as a link-prediction problem and utilise set of Random Forest predictors for the prediction task. We use data collected by the Copenhagen network study; a study unique in scope and scale and tracks 847 students via mobile phones over the course of a whole academic year. We find that network and social features hold the highest discriminatory power for predicting future encounters.
... Proximity was assessed by measuring the Received Signal Strength (RSSI) of Bluetooth signals from nearby devices: a high RSSI means that the two devices are physically close, a low measure indicates that devices are further apart or that there are obstacles in between. 49 The OD dataset was collected in an office workplace staffed with about 200 people that agreed to wear for several days wearable proximity sensors while in the workplace; 46, 50 these sensors used low-power radio communication as a proxy for the close-range proximity of individuals wearing the devices. 35 Both the SD and OD datasets represent close-range proximity interactions as a temporally-ordered sequence of contact networks, where nodes are individuals and an edge between nodes indicates a close-range proximity relation. ...
... The devices measured and recorded the Received Signal Strength (RSSI): a high RSSI means that the two devices are physically close, a low measure indicates that devices are further apart or that there are obstacles in between. 49 • The OD dataset was collected by the SocioPatterns collaboration, using an infrastructure based on wearable sensors that exchange radio packets, detecting close proximity (≤ 1.5m) of individuals wearing the devices. 35 The temporal resolution of the data is of 20 seconds. ...
Article
Non-pharmaceutical interventions are crucial to mitigate the COVID-19 pandemic and contain re-emergence phenomena. Targeted measures such as case isolation and contact tracing can alleviate the societal cost of lock-downs by containing the spread where and when it occurs. To assess the relative and combined impact of manual contact tracing (MCT) and digital (app-based) contact tracing, we feed a compartmental model for COVID-19 with high-resolution datasets describing contacts between individuals in several contexts. We show that the benefit (epidemic size reduction) is generically linear in the fraction of contacts recalled during MCT and quadratic in the app adoption, with no threshold effect. The cost (number of quarantines) versus benefit curve has a characteristic parabolic shape, independent of the type of tracing, with a potentially high benefit and low cost if app adoption and MCT efficiency are high enough. Benefits are higher and the cost lower if the epidemic reproductive number is lower, showing the importance of combining tracing with additional mitigation measures. The observed phenomenology is qualitatively robust across datasets and parameters. We moreover obtain analytically similar results on simplified models.
... A number of large-scale data collection experiments have already been performed to overcome these limitations. These studies make use of modern social sensing technologies to track the proximity of individuals, a good proxy for measuring social-group-interactions. For example, physical proximity between mobile phone users-detected via Bluetooth signal strength-has been used to measure the strength of friendship ties [410] and to unveil the structural patterns in longitudinal data sets [411]. Although these data, released by the Copenhagen Networks Study [412], would represent a good testing ground for higher-order representational algorithms, they miss one essential component, that is information about social phenomena unfolding on top of the social structure. ...
Thesis
Full-text available
Complex networks have been successfully used to describe the social structure on top of which many real-world social processes take place. In this thesis, I focus on the development of network models that aim at capturing the fundamental mechanisms behind the dynamics of adoption of ideas, behaviours, or items. I start considering the transmission of a single idea from one individual to another, in an epidemic-like fashion. Recent evidence has shown that mechanisms of complex contagion can effectively capture the fundamental rules of social reinforcement and peer pressure proper of social systems. Along this line, I propose a model of complex recovery in which the social influence mechanism acts on the recovery rule rather than on the infection one, leading to explosive behaviours. Yet, in human communication, interactions can occur in groups. I thus expand the pairwise representation given by graphs using simplicial complexes instead. I develop a model of simplicial contagion, showing how the inclusion of these higher-order interactions can dramatically alter the spreading dynamics. I then consider an individual and model the dynamics of discovery as paths of sequential adoptions, with the first visit of an idea representing a novelty. Starting from the empirically observed dynamics of correlated novelties, according to which one discovery leads to another, I develop a model of biased random walks in which the exploration of the interlinked space of possible discoveries has the byproduct of influencing also the strengths of their connections. Balancing exploration and exploitation, the model reproduces the basic footprints of real-world innovation processes. Nevertheless, people do not live and work in isolation, and social ties can shape their behaviours. Thus, I consider interacting discovery processes to investigate how social interactions contribute to the collective emergence of new ideas and teamwork, and explorers can exploit opportunities coming from their social contacts.
... where time t is in units of days and s ij (t) is the maximum Bluetooth signal strength between participants i and j measured during day t. The threshold 80 dBm corresponds to a distance of about 2 m and maximises the ratio of social interactions to transient and unimportant connections [61]. In order to minimise noise from the beginning and end periods of data collection, in this study we focus on the period from the first of February 2014 to the end of April 2014, which corresponds to the spring semester and is in the middle of the "SensibleDTU 2013" data collection, the second deployment of CNS. ...
Preprint
Full-text available
Spreading or complex contagion processes on networks are an important mechanistic foundation of tipping dynamics and other nonlinear phenomena in complex social, ecological and technological systems. Increasing amounts of temporal network data are now becoming available to study such spreading processes of behaviours, opinions, ideas, diseases, innovations or technologies and to test hypotheses regarding their specific properties. To this end, we here present a methodology based on dose-response functions and hypothesis testing using surrogate data sets. We demonstrate this methodology for synthetic temporal network data generated by the adaptive voter model. Furthermore, we apply it to empirical temporal network data from the Copenhagen Networks Study. This data set provides a physically-close-contact network between university students participating in the study over the course of three months. We study the potential spreading dynamics of the health-related behaviour "regularly going to the fitness studio" on this network. Based on a hierarchy of surrogate data models, we find that the empirical data neither provide significant evidence for an influence of a dose-response-type network spreading process, nor significant evidence for homophily. The empirical dynamics in exercise behaviour are likely better described by individual features such as the disposition towards the behaviour, and the persistence to maintain it, as well as external influences affecting the whole group, and the non-trivial network structure. The proposed methodology is generic and promising also for applications to other data sets and traits of interest.