A simulation-based approach to evaluating the reliability of using Bluetooth signal strength to detect parent-child interactions

Early in the development of the Open Dynamic Interaction Network (ODIN) platform for responsive Ecological Momentary Assessment (rEMA), we aimed to detect physically co-located social interactions using Bluetooth signal strength. The goal was to estimate the distance between a child (having a beacon) and their parent (having a phone). Our early initial pilots were marked by ad-hoc selection of signal strength thresholds (e.g., considering received signal strength of -70 to indicate “interaction”). However, reproducibility and validity tests with members of our research revealed variations in our ability to detect interaction as a static state correctly. This early experience pointed to a need for a more systematic framework for detecting the beginning and ending of interaction periods. Ideally such a framework should be able to incorporate researcher-specified parameters such as the expected physical separation during interactions. It should also be capable of being readily recalibrated to new physical environments with different electromagnetic profiles or adapted to new phone technologies as they become available. Finally, it should allow for robust estimation of type I and type II errors in detecting the start and end of interactions. This paper describes such a framework.