From Sensors to Insights: Measurement Quality Considerations for Sensor Studies in Everyday Contexts

Abstract

Consumer-grade mobile and wearable sensors can give insights into people’s behavior and affect in everyday-life settings. However, the heterogeneous data collection settings require particular attention to data quality. We identify relevant quality dimensions for sensor data collected in such studies based on social-science focused and technology-focused data quality frameworks. With a systematic literature review of field studies applying geolocation, heart rate, and acceleration sensors, we then compare the quality considerations from the literature with those reported in actual studies. From this, we derive a set of reflection prompts for researchers working with consumer-grade sensor data outside of lab contexts. Notably, we encourage researchers to consider and report on the discriminant validity of their measures, add real-time data quality warnings, and motivate decisions like feature selection and imputation methods. Thus, we contribute to improving the internal and external validity of technology-supported studies of humans and their behavior.