Passive Social Sensing with Smartphone: A Systematic Review (Preprint)

BACKGROUND Smartphones are widely used and have become hubs of personal communication. Combined with multiple sensors, they are capable of monitoring social interaction passively. OBJECTIVE To review the published empirical English literature of social sensing with smartphones passively. METHODS Following the PRISMA guidelines, we constructed a search string considering variations of the term smartphone. The search was conducted in ACM Digitial Library, IEEE Xplore and Web of Science. Snowballing method was also applied following Wohlin guidelines after the initial search. Papers were included if they were empirical, only used sensors on smartphones to collect data, measuring social interaction or social activity level, required minimum user interaction and described as passive. RESULTS The search produced 688 results, of which 15 eligible articles were identified, two more articles were added from the snowballing. Participants of included studies range from 5 to 163, and described experiments length range from two weeks to two years. The aim of all studies is to understand human social behaviours, especially, some of them correlate social interaction levels with health-related topics. Android is the most popular operating system for capturing calls, messages and Bluetooth data, which are the three most frequently used sensors. Various data analysis methods, including simple as correlation, complex as machine learning, have been utilized. Benefits such as ubiquitousness, unobtrusiveness, powerfulness, and continuity and challenges such as privacy, accuracy and methodology were reported and summarized from reviewed studies. CONCLUSIONS Although the reviewed studies confirmed the feasibility and validity of passive smartphone social sensing, they suffered from issues of sensing strategy, privacy concerns, sample sizes, the significance of experiments and data integrity. The recommendation is to conduct more research on a theoretical basis, build personalized models and implement state-of-art technologies. The smartphone social sensing technology provides innovative opportunities to measure human social behaviour objectively. It has a promising future with concerns including methodology, privacy and evolvement addressed by interdisciplinary collaborations between technology experts, computer scientists and psychologists.