High-Throughput Digital Phenotyping

Social, behavioral, and cognitive phenotypes have traditionally been difficult to study due to their temporal and contextual dependence. The standard approach to learn about, say, behavior is to administer surveys, which are cross-sectional, subjective, and often burdensome to subjects. We believe that the ubiquity and capability of smartphones–when coupled with appropriate data analytic techniques–can be part of the solution. We have coined the term Digital Phenotyping to refer to the “moment-by-moment quantification of the individual-level human phenotype in situ using data from personal digital devices, in particular smartphones.” Unlike the large majority of commercially available smartphone applications, Beiwe is intended for use in research. Most applications, out of convenience, use software development kits (SDKs) made by Apple and Google which usually generate unvalidated behavioral summary measures using closed proprietary algorithms that do not meet the high standards of reproducible science. In contrast, Beiwe collects raw sensor and phone use data and Forest analyzes these raw data. That means that data collection and data analysis are customized to address specific scientific questions of interest rather than modifying scientific questions based on what data happens to be available.