Riskalyzer

Abstract

An individual's risk-taking propensity is "the stable tendency to choose options with a lower probability of success, but greater rewards". This risk propensity plays a central role in decision making by customers as well as managers, and is a mediator in behavior associated with security, privacy, health, finance, and well-being. Most common approach to understanding an individual's risk propensity remain lab-based games and surveys. Administering such surveys and games is a manual, time, and money intensive process that is also fraught with multiple biases. Recently, smartphones are increasingly seen as large-scale sensors of human activity, recording data related to physical and social aspects of people's lives. Building on this trend we investigate the potential of passive phone-based data for automatically inferring an individual's risk propensity. Specifically, we describe a novel approach to model an individual's risk propensity based on her mobile phone usage. Based on a 10-week field + lab study involving 50 participants, we report that: (1) multiple phone-based features (e.g., average gyradius) are intricately associated with participants' risk propensity; and (2) a phone-based model outperforms demography-based models by 39% in terms of accuracy of predicting risk propensity. In organizational terms, a better understanding of risk behavior could contribute significantly to risk management programs. At the same time, such results could open doors for more nuanced understanding of the underlying human risk phenomena and their interconnections with social and mobility behavior.