Physiological Computing

Development of a mobile App to improve the Understanding of Flow by Knowledge Workers

Flow is defined as a state in which people are completely absorbed in their activity. Since increased flow is associated with both improved performance and an enhanced sense of well-being, this state is considered particularly desirable in the work context. Therefore, in this project, we developed a mobile flow app for knowledge workers that enables app users to better understand their flow throughout the day (e.g., when productive phases occur). The developed flow app collects physiological data, classifies flow states using machine learning methods, and presents the flow classification results aggregated and visualized in the app in addition to the physiological data (e.g., heart rate).

This figure shows a screenshot of the “Analytics” tab of the developed Flow app. A bar chart can be seen that gives users an insight into their flow experience. One bar shows the percentage of flow of the respective user in relation to the recorded data for each hour of the selected day.

Development of an adaptive Instant Messaging Blocker to maintain Flow during productive Periods

This figure illustrates the adaptation logic of the Instant Messenger Blocker. Based on the measured physiological data of a person, flow detection is performed with the help of machine learning methods. If a flow is detected, the “Do-not-Disturb” status of the Instant Messenger is activated for the flow duration. If no flow is detected, the “Do-not-Disturb” status is deactivated.

Notifications from instant messaging applications can interrupt employees' productive periods. While there are manual ways to influence instant messenger notification behavior, such as turning off the application or muting notifications for certain periods of time, these measures require self-discipline and often result in missed notifications when not in flow. The developed adaptive instant messaging blocker aims to solve this problem by detecting users' flow state at predefined intervals, based on their physiological data and with the help of machine learning methods. Once a flow state is detected, the Do-not-Disturb status is automatically activated for the duration of the flow state.

Classification of Flow based on Heart Rate Variability using Machine Learning

This figure presents an overview of the process used to generate the training data for machine learning. The physiological data serves as the input, and the survey data serves as the output. Based on the timestamps of the data collection, the respective physiological data are mapped to the corresponding survey data. On this basis, the training of the machine learning model can take place. In our example, we are concerned with a classification problem that can distinguish between flow and no flow.

Flow is considered a desirable state because it is related to improved performance and well-being. However, current methods cannot measure a person's flow continuously or without interrupting current activity. In this project, flow was classified by machine learning using heart rate variability (HRV) data collected in a field study with knowledge workers. A classifier was trained here to distinguish between "flow" and "no flow." This classification provides a foundation for the development of adaptive systems that can continuously detect flow without interruption and intervene on that basis.

Developing an Infrastructure for Researching Students’ Flow in Remote Learning

On the left side of the figure, the developed infrastructure is shown. The infrastructure was developed based on a combination of a Raspberry Pi mini-computer and a Polar H10 chest belt wearable to collect physiological data. We used “Raspbian” as the OS. An SQL database is used to securely store all data recorded by our Python application. A second application accesses the database to present the recorded data in a web dashboard. This dashboard is served by a Python web server running on the infrastructure. Users can connect to a WiFi hotspot opened by the infrastructure to access the web dashboard using their individual login credentials. On the right hand side of the figure, the experience sampling method (ESM) browser extension is shown. We implemented a logic for the well-established ESM as we also wanted to collect survey data. In ESM, participants are surveyed several times throughout the day for the defined study period. Specifically, the ESM logic was implemented in a browser extension for Google Chrome. In addition, the browser extension is also able to collect browsing behavior data for predefined URLs from a whitelist. Hereby, the browsing behavior and survey data are stored in a secure online database. The ESM browser extension is decoupled from the rest of the infrastructure and runs independent from the users' computer system.

In this project, we developed an infrastructure to collect physiological data, surveys, and browsing behavior data to capture students’ learning experience during remote learning. Our solution is based on a Raspberry Pi minicomputer and a Polar H10 chest strap. Preliminary results and experiences from a field study with medical students not only show the potential to collect different types of data during remote learning, but can also serve as a basis for future work. For example, a less intrusive and continuous measurement method can be developed based on classifying cognitive-affective states such as flow or other learning-related constructs with the collected data using machine learning.

By car

By bus, subway and train

The H-Bahn (Suspended Monorail System)

By airplane

At campus

Navigation and maps