Reliability of Bluetooth-based connectivity traces for the characterization of human interaction

The characterization of human interaction at different levels has been a matter of interest in many disciplines. So far, social networking through the Internet has been the main source to infer human beings’ relations. Nevertheless, due to the irruption of wearable devices with wireless communication capabilities, initiatives that use them to measure physical proximity are increasingly popular. Built-in wireless communication technologies allow these devices to detect each other and to infer their owners’ interaction, based on their proximity measurements. This approach, which is followed by most proximity initiatives in the research community, poses three main challenges that usually limit the quality of collected data and consequently, the reliability of human behavior characterization: the person-device uncertainty, the sample period and the bias caused by the particularities of the underlying wireless technology. The work presented here analyzes empirically the impact of these three limitations when Bluetooth is the communication technology used to detect proximity. It also presents the expansion of the results when additional mechanisms to counteract the impediments are applied, and it states their necessity for the reliability of the results. They show relevant differences with previous initiatives that open a discussion on the proper use of wireless wearable devices as a tool for the characterization of human interactions.