The iLog methodology for fostering valid and reliable Big Thick Data

Nowadays, the apparent promise of Big Data is that of being able to understand in real-time people's behavior in their daily lives. However, as big as these data are, many useful variables describing the person's context (e.g., where she is, with whom she is, what she is doing, and her feelings and emotions) are still unavailable. Therefore, people are, at best, thinly described. A former solution is to collect Big Thick Data via blending techniques, combining sensor data sources with high-quality ethnographic data, to generate a dense representation of the person's context. As attractive as the proposal is, the approach is difficult to integrate into research paradigms dealing with Big Data, given the high cost of data collection, integration, and the expertise needed to manage them. Starting from a quantified approach to Big Thick Data, based on the notion of situational context, this thesis proposes a methodology, to design, collect, and prepare reliable and valid quantified Big Thick Data for the purposes of their reuse. Furthermore, the methodology is supported by a set of services to foster its replicability. The methodology has been applied in 4 case studies involving many domain experts and 10,000+ participants from 10 countries. The diverse applications of the methodology and the reuse of the data for multiple applications demonstrate its inner validity and reliability.