Staying afloat in the home-office

At IDUN, we also had revisit our digital workplace infrastructure, habits and interaction with each other. By now, we found our rhythm in digital meetings, are comfortable doing all project management on the cloud, brainstorm on digital whiteboards and even have remote beers in the evening (although we miss the real ones dearly). We even hired three interns that are working on a different continent, something we would a deemed impossible two years in the past. And just like many other companies, we found that we can work just as well together and without losing momentum in reaching our ambitious goals.

At the same time, we think about how our technology could be used to tackle some of the mental challenges associated with working mostly remotely. Using neurofeedback, we imagine that productivity software such as Microsoft Office could assist the user’s wellbeing by suggesting small breaks or mindfulness sessions in the calendar, playing relaxing music in times of stress or by connecting us with friends and colleagues. Similar to other health metrics such as activity or heart rate, a personal assessment of stress levels, focus time and mental load over the course of a week could be of high interest for individuals interested in the quantified self. Next to the workplace at home, these metrics could additionally interface with our smart appliances or favorite apps, to make them more emphatic. Neurotech could help us take better care of ourselves.

To reach that end, some major challenges need to be tackled, one of the most important ones being data privacy and ownership of the data by the user. With sensitive biodata being recorded and processed in a cloud environment, a zero trust architecture [6] where every access request validated based on user identity, device, location, and other variables, seems to be a necessity. If user can be identified by the data recorded (PII), a data architecture is required that allows users to track, access and delete their data at any given time. What this means for machine learning models, that have been trained on that user data, is an ongoing debate. Differential privacy, or introducing artificial ‘noise’ to datasets is an option to prevent reverse identification at the cost of model accuracy. With increasing processing power and bandwidth on the edge, another promising route to ensure data privacy is the use of federated learning, where classifiers are trained decentralized on the user device and no actual user data is exchanged [7].