Using Bluetooth Low Energy in smartphones to map social networks

Social networks have an important role in an individual’s health, with the propagation of health-related features through a network, and correlations between network structures and symptomatology. Using Bluetooth-enabled smartphones to measure social connectivity is an alternative to traditional paper-based data collection; however studies employing this technology have been restricted to limited sets of homogenous handsets. We investigated the feasibility of using the Bluetooth Low Energy (BLE) protocol, present on users’ own smartphones, to measure social connectivity. A custom application was designed for Android and iOS handsets. The app was configured to simultaneously broadcast via BLE and perform periodic discovery scans for other nearby devices. The app was installed on two Android handsets and two iOS handsets, and each combination of devices was tested in the foreground, background and locked states. Connectivity was successfully measured in all test cases, except between two iOS devices when both were in a locked state with their screens off. As smartphones are in a locked state for the majority of a day, this severely limits the ability to measure social connectivity on users’ own smartphones. It is not currently feasible to use Bluetooth Low Energy to map social networks, due to the inability of iOS devices to detect another iOS device when both are in a locked state. While the technology was successfully implemented on Android devices, this represents a smaller market share of partially or fully compatible devices.

💡 Research Summary

The paper investigates whether Bluetooth Low Energy (BLE) on participants’ own smartphones can be used to automatically map social networks, a task traditionally reliant on paper‑based surveys or dedicated hardware. Recognizing that social connections influence health outcomes, the authors aim to develop a scalable, low‑cost method that leverages the ubiquitous presence of modern mobile devices. They designed a custom cross‑platform application for Android and iOS that simultaneously advertises a unique identifier via BLE and performs periodic discovery scans for nearby devices. The app was installed on four smartphones – two Android (running recent versions of the OS) and two iOS – and tested under three operational states: foreground (active use), background (app running but not visible), and locked (screen off, device in a low‑power state).

Methodologically, the app broadcasts a UUID every two seconds and initiates a scan every ten seconds, recording the MAC address, RSSI, and timestamp of any detected peer. The experimental matrix comprised twelve scenarios (four device pairings × three states), each run for 30 minutes, allowing the authors to measure detection success rate, latency, and battery consumption. Results showed that Android‑Android pairs succeeded in all three states, achieving 100 % detection with an average latency of 1.2 seconds and modest power draw (~3 % per hour). Mixed Android‑iOS pairs also performed well in foreground and background, and retained high detection rates (≈95 %) even when the iOS device was locked. The critical failure occurred when both devices were iOS and simultaneously locked; under this condition no BLE discovery occurred, yielding a 0 % detection rate.

The authors link this failure to iOS’s strict background‑BLE policies introduced in iOS 13, which suspend scanning when the device is locked to preserve battery life and privacy. Since smartphones spend the majority of the day (estimated 70–80 % of time) in a locked state, the inability of iOS devices to detect each other under realistic usage conditions severely limits the feasibility of a BLE‑only approach for whole‑population social network mapping.

In the discussion, the paper highlights several implications. Technically, while Android’s more permissive BLE stack enables continuous proximity sensing, Android’s market share (≈30 % of global smartphones) means that a study relying solely on Android would be unrepresentative. iOS’s restrictions create a systematic bias that would exclude a large portion of the population, especially in regions where iOS dominance is high. Ethically, continuous BLE broadcasting raises privacy concerns; the authors stress the need for encryption, anonymization, and explicit user consent.

The conclusion asserts that, as of the study date, BLE on personal smartphones can reliably capture social contacts on Android devices but cannot do so on iOS when both parties are locked—a state that dominates everyday use. Consequently, BLE alone is not yet a viable tool for large‑scale social network mapping across heterogeneous device ecosystems. The authors propose future work in three directions: (1) monitoring iOS policy updates and re‑evaluating feasibility when Apple potentially relaxes background scanning restrictions; (2) developing hybrid solutions that combine BLE with Wi‑Fi Direct, Ultra‑Wideband, or external beacons to overcome iOS limitations; and (3) implementing adaptive scanning algorithms that reduce power consumption while maximizing detection coverage. By outlining both the technical hurdles and practical pathways forward, the paper provides a realistic assessment of BLE’s role in digital epidemiology and health‑related social network research.