Adaptive Mobile Sensing: Leveraging Machine Learning for Efficient Human Behavior Modeling; Social Challenges of Uber’s Driver Rating System

The purpose of the technical project is to provide a framework for how smartphone sensor data can be collected, cleaned, stored, and modeled to effectively predict human states. Outlined in the paper is a three-week long study involving an application named Sensus, which collects and aggregates user smartphone data. Differing models for how the application collects data were tested in order to determine which strategy collects data of the highest quality with minimal battery consumption. Sensing strategies range from infrequent sampling to continuous sampling. A final, dynamic data collection strategy was then explored, using a machine learning model trained on existing data collected from 220 participants to determine the best times to trigger sensor sampling. Results of this study will include 1) extraction of model features that deliver maximized data quality with minimized battery consumption as compared to pre-existing baseline models, 2) implementation of context-driven modeling of user smartphone data on user’s contextual environment, and 3) customization of a time-series database for optimized data queries used in metadata visualizations. The adaptive sensing models produced could be used in future large population studies that efficiently examine patterns of behavior in multiple individuals over extended periods to identify disease indicators present in an average user’s daily life.

The purpose of the STS research is to explore the social ramifications of Uber’s star rating system on Uber’s workforce-user base and the schism it creates in its labor-market segment. This research specifically seeks to answer how varying social groups have influenced the use of Uber’s driver rating system and its subsequent effects on drivers. Methods of network analysis and wicked problem analysis are used to reconstruct the complexity of the issues faced between social groups, namely that of social bias, and subsequently explore their connections to the rating system and the larger Uber environment. The Social Construction of Technology framework is used to further explore how these engrained societal viewpoints of varying driver segments affect the use of the rating system. Findings from this research suggest that the racial biases of riders and the lean application structure of Uber plays into creating an asymmetrical information space that leads passengers to rate their riding experience based on comparison to (uncommunicated) expectations as opposed to the objective quality of provided service itself. The resultant network structure centralizes power in part-time non-immigrant drivers, which, inadvertently, allows them easier access to increased ratings over the long run. This research is important in learning how to generate social rating systems on larger scales that avoid unnecessary subjectivity or account for it such that specific user groups are not disenfranchised by future systems.

School of Engineering and Applied Science
Bachelor of Science in Systems and Information Engineering
Technical Advisor: Laura Barnes
STS Advisor: Bryn Seabrook
Technical Team Members: Erin K. Barrett, Cameron M. Fard, Hannah N. Katinas, Charles V. Moens, Lauren E. Perry, Blake E. Ruddy, Shalin D. Shah, Tucker J. Wilson