My research is motivated by the need to solve large-scale collective-action problems: how do we live in cities with little or no pollution?; Avoid infectious disease outbreaks?; Decrease traffic congestion? These challenges require us to confront a fundamental tension:
how do we reconcile an individual’s desire for agency and choice in what they do, with what everyone wants for society as a whole?
The goal of my research is to understand human behavior at scale, and to empower individual decision making. To do so, I develop innovative algorithms, build systems and perform experiments.
Theories of collective action, bounded rationality, and persuasion speak to this issue, but they do not take into account technology use (ubiquitous cellphones) or grassroots advertising. Work on collective action (e.g., Ostrom) focused on how small (typically ~100; largest being ~10,000), homogenous communities, with the ability to monitor at low-cost and to sanction, can cooperate to overcome the ‘tragedy of the commons.’ The question of scaling to large, heterogeneous populations is left open. While work on bounded rationality (e.g., Kahneman) analyzed the role of systematic cognitive biases in individual decisions and work on persuasion (e.g., Cialdini) analyzed the effects of norms on individual behavior, the latter two theories do not address how individual decisions, which are influenced by cognitive biases and social norms, explain collective behavior.
My research agenda focuses on how computing can bridge the gap between theories of individual behavior and collective action. Three realizations underpin my agenda. First, large-scale technological networks were helping satisfy Ostrom's necessary condition for cooperation---low-cost communication and monitoring; second, networks, through their algorithms and incentive structures mediate our observations of behavior---shaping norms and influencing interpretation; third, computing (i.e., large-scale data analysis; algorithmic synthesis of messages) could facilitate a scaling effect on theories of collective action. Thus, I began to develop models of behavior that address scale and heterogeneity; develop mechanisms that improve social-welfare; design algorithms for synthesis of messages that incorporate scholarship on cognitive biases and the role of norms.
My current research includes work on understanding large-scale behavior (community discovery;discovery of coordinated behavior;sampling networks; robust neural models; strategic interaction), connecting individual behavior to macroscopic patterns (growth of networks; network failures; voting); and in building systems (experiments in the wild; localization; preserving privacy).
Building sustainable, large-scale networks to support collective action challenges (e.g., health) is non-trivial, since individuals are invested in using their current social networks and information services. These popular platforms offer free services and are often not a neutral party to the information exchange—they have a clear commercial motive (e.g., advertising) that may run counter to an individual's desire (e.g., living a healthy lifestyle). How do we address this tension that arises due to the mis-alignment of incentives between the technological platform and the individual? Since individuals cannot easily abandon their use of the platform due to network effects (e.g., all their friends are on that platform), there exists a power imbalance. Some of our ongoing work addresses these issues in the context of online advertising.
As computing artifacts become increasingly intertwined with our everyday, seemingly neutral services and platforms subtly guide our behavior. Against this backdrop, it is increasingly important to understand the broad behavioral patterns that emerge and help individuals break free from the subtle control. My research contributes to this mission by developing sophisticated algorithms for identifying patterns at scale, understanding how knowledge of these patterns influence individual decisions and conversely how individual decisions result in the observed patterns, and developing systems to support informed decision-making.