DIAG Social Experiment

The DIAG experiment was deployed at the Department of Computer, Control, and Management Engineering “Antonio Ruberti” at Sapienza, University of Rome. It lasted around 5 days, from October 17, 2011 to October 21, 2011, and involved 116 participants, including undergraduates, graduates and Ph.D. students.

We asked each of 116 students to wear an active OpenBeacon RFID tags which periodically sent information about the approximate location of the wearing person (i.e., it depends if the reader had received the packet) and its social contacts (i.e., if the person was close to any other person). An example information we got can be tag 1274 met tag 1055 in room 3, or tag 1143 is close to room 5.

Also, we stored additional information for each person participating to the experiment, which was the following:

– Age
– Sex
– Course of study
– Academic year

This only allowed us to observe how people clustered; no other information was needed. Also, we didn’t store anything about their identity, such as name or ID. It should be clear that from this generic information it’s not possible to identify who is the owner of a specific tag.

Messages sent by tags were collected by OpenBeacon RFID readers and forwarded through the ethernet to a central collector server, where data were processed offline. At the end of the experiment we collected about 200 MB of data log; then we parsed them to analyze the relationships among students and understand their sociability.

In order to represent and visualize our graph, we used Gephi. The following figure depicts social interactions among students during our social experiment.

The graph was initially represented by a weighted adjacency matrix in order to show only node IDs and their interactions; then, we built a GEXF graph by using our gexf4j library, so as to visualize other information, such as age, sex and so on. Size and color of a node shows the number of neighbors for each individual (i.e., its degree), while thickness of an edge tell us how long two individuals interacted. Specifically, yellow and small nodes have few relationships with others, while red and big nodes are very sociable students.

If you use this data set, please use the following citation and link this page in your work:

Luca Becchetti, Lorenzo Bergamini, Francesco Ficarola, and Andrea Vitaletti. 2012. Population protocols on real social networks. In Proceedings of the 9th ACM symposium on Performance evaluation of wireless ad hoc, sensor, and ubiquitous networks (PE-WASUN ’12). ACM, New York, NY, USA, 17-24. DOI=10.1145/2387027.2387031 http://doi.acm.org/10.1145/2387027.2387031

Below you can find all data collected by the experiment. CSV, GEXF, DNF and other output graph formats can be obtained by using OpenBeaconParser, a software I’ve developed to parse raw data log. OpenBeaconParser can be downloaded by everyone and you can also obtain other kinds of formats (please consult the README file and the software help).

Social graph (vector, pdf)
Raw data log
SQL database (mysql)
Aggregated graph (csv, weighted adjacency matrix)
Aggregated graph (gexf)
Dynamic graph (dnf)
Dynamic graph with SQL data (json)

<<< DOWNLOAD ALL FILES (tar.bz2) >>>

In the raw data log you can find two kinds of packets: S or C. S packets are related to sightings, namely when an individual is close to a RFID reader, while C packets are interactions. In both packets the common fields are:

t : timestamp
ip : reader IP in HEX
id : tag ID who sent the packet
seq : a unique sequence number for each packet

Additionally, C packets include a target node ID who was in the interaction. For instance, in the string [1119(1) #1] the target node ID is 1119. All other fields are for debugging.

If you like observing what are the interactions among participants over time in a nice live application, please click on the following image.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.