Scalable Detection Of Perceived Stress-Zakaria
#notesFromPaper Year : Tags : Authors: Zakaria Balan Lee
Camellia Zakaria, Rajesh Balan, and Youngki Lee. 2019. StressMon: Scalable Detection of Perceived Stress and Depression Using Passive Sensing of Changes in Work Routines and Group Interactions. Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 37 (November 2019), 29 pages. DOI:https://doi.org/10.1145/3359139
stress is incredibly debilitating to productivity and general life satisfaction, with stress being able to manifest in physical ailments. The paper mentions that the WHO estimates a 1 trillion dollar loss to the economy caused by depression. Some of the problems with standard stress sensing measures are that they require users to install dedicated apps or even physical sensors. The authors of this paper developed a program called “StressMon” which uses wifi infrastructure to detect movement patterns and location data. The authors focused on physical group interactions since workplace stress is often caused more by group dynamics than individual work. Supportive groups allow individuals to receive peer support, which dramatically reduces stress levels. StressMon works using a server-side application that tracks user anonymised MAC IDs (this avoids forcing the user to install the app on their device). By seeing the number of users connected to the same router (in the same room), group sizes could be determined. The study was longitudinal with hundreds of participants. Participants came from a variety of majors, including Computer Science, Finance, and Accounting. Pre and post study questionnaires determined stress levels. StressMon’s #machineLearning algorithm was able to identify depression with 88% accuracy, a better accuracy than many dedicated apps with finer grained control. There were a variety of limitations with this application. First and foremost is that the application works in indoor locations with wifi mesh networks amenable to StressMon. In addition, the group dynamics of a university (where this study took place) are very different to other locations, e.g. a hospital or corporate setting. While StressMon had informed consent from the members it was tracking, this may not be the case by unscrupulous users of StressMon, who install it within their offices without informing participants.
This paper is not directly about distributed cognition. Despite this, I still think this paper is a fantastic application of distributed computing since it recognizes that the major problem in workspaces are group dynamics and interactivity, instead of the simplistic “person in front of computer” paradigm that HCI often focuses on. Future refinement of this application (either via additions to the server-side application or dedicated companion apps installed to the users phone) could be extremely beneficial in determining more granular stress relationships. Perhaps a person feels differently about certain groups than others. Perhaps a person has differing stress levels throughout the day. Tracking all of this (in an ethical manner) would make group dynamics so much better for all workers.
