Visualizing Human Interactions in a Workspace Setting and Maintaining Privacy

To better understand how humans interact in office setting, it is necessary to provide sufficiently understandable tooling for interaction analysis. At the same time there is a requirement to maintain the balance between privacy and utility. Using interaction measurement framework developed by the MIT, we utilized network analysis for determining an approximation of location, voice activity detection (VAD) for determining discussions, and spatio-temporal analysis (STA) for creating a visualization. We aimed to provide tooling for evasive analysis of workplace interactions, which would be intuitive enough to be understandable by non-technically oriented actors. Here we detail implementation details and decision reasoning.