Primary With Daniel Thomas. PNAS 2023
Measured the effect of network-level disruptions of 6 hate organizations on their former audiences on Facebook, finding that more peripheral members of the audience reduced their consumption and production of hate content, while there was a temporary backlash among users who were closest to the de-platformed users, but that this backlash subsided within two months. Used a staggered diff-in-diff design to estimate effects.
Primary with Lisa Singh and Yanchen Wang. Also with Yifang Wei, Christo Kirov, Susan Martin, Katharine Donato, Yaguang Liu and Kornraphop Kawintiranon.
Integrating publicly available organic data from social media and newspapers with more traditional indicators of forced migration can improve performance of models to determine when and where people will move, better enabling the pre-allocation of aid resources. Used Bayesian hierarchical models to blend data at different spatial and temporal granularity to model displacement in Iraq.
With Nick Hagar and Eric Dunford
We propose a novel, scalable method for generating individual-level influence measures across a set of social media cascades, and demonstrate its applicability in real scaled data.
With Garren Gaut, Andrea Navarette, Paul van der Boor, Adolfo de Unánue, Jorge Díaz, Eduardo Clark, and Rayid Ghani,
Designed automation system for processing citizen petition for the Government of Mexico, using natural language processing and machine learning techniques.
Reports and Articles
December 2016, Global Policy Journal
Report from a year-long international foresight summit, I participated as US delegate on the transnational terrorism working group for the Global Governance Futures Group by Bosch Foundation. Authored with Aryaman Bhatnagar, Elisa D. Lux, Yuan Ma, Minako Manome, Sarah Markiewicz, and Fanglu Sun.
Conducted with the Institute for Defense Analyses.
Developed a probabilistic model to predict future US ground force demand using historical data. Part of effort to hedge US armed forces against an uncertain future, and explore the effects of alternative force mixes and readiness strategies. Interfaces with a cost model to provide twenty year estimates under different expectations about the future, or force generation policies. Published in several IDA reports. (For example, see also the Air Force Cost Model)
Conducted with the Institute for Defense Analyses
Used process tracing to explore the ways in which in-year changes in fuel price could affect readiness, conducted interviews with relevant parties in OSD and the services. Concluded that fuel volatility has not had an effect on readiness, but that it represents an exogenous shock that could multiply extant budgetary difficulties. Published in IDA report.
Meso: A Structure for Networked Governance
With Anna Waldman-Brown, Salman Aldukheil. Semi-finalist proposal for the Global Challenges Foundation.
Meso is a networked governance structure designed to fill the governance gap between state-level international organizations, municipal governments and community organizers. Meso will tackle global issues by encouraging locally appropriate action at the local-level. Meso will accomplish this by facilitating formation of global innovation networks, and the flow of focused information across them, bringing critical information to those who need it.
Violent groups must continue to adapt in order to compete with rivals and stay one step ahead of law enforcement, but adopting new tactics is risky because experimentation and failure are costly. I argue that violent organizations ease the costs of adopting new techniques by learning from others. However, this learning is restricted because sharing risks losing operational information to rivals or the state. Groups therefore safeguard tactical information, sharing only with trusted partners. This tension is exacerbated in competitive environments.
Using data on violent group networks and attacks, I support this theory with three findings: (1) Groups are more likely to adopt tactics previously used by trusted partners. (2) Tactics diffuse more readily in networks that facilitate information flow. (3) Groups with rivals are more likely to adopt tactics from their partners, but do not imitate rivals despite similar environments.
Anti-Social Networks: The Effects of Violent Group Cooperative Networks Structure on Capacity for Violence, and Survival and Diffusion in Networks of Violent Groups
Most militant groups do not act in isolation. They exist in a vast web of partnerships which help them to mobilize resources and survive. How does the pattern of partnerships in which a group is embedded shape the ways in which the group can take advantage of those resources, and the ways in which they are made vulnerable by the network?
I argue that both the structure of a local network of partnerships in which a group is embedded, and a group’s position in it, shape the extent to which that group benefits from that network. Some local networks center around one or a few well-connected groups. In these local networks, those influential groups have leverage because they can control the flow of goods and information across the network. Influential groups specialize in coordination, helping partners to survive and increasing their capacity for violence. Other local networks are flatter, with no groups dominating. These local networks lack coordination, are less capable of screening out vulnerable partners, and thus are more vulnerable to security forces. As a result, members of these communities, even relatively influential ones, have lower capacity for violence. Better connected groups in decentralized networks are actually more vulnerable, and do not survive as long.
In this paper, I develop a theory to explore (1) the tradeoffs that transregional terrorist groups make when deciding whether to intervene in a local war, and which group to sponsor; (2) the tradeoffs made by the sponsored groups when deciding whether to accept or solicit sponsorship from the transregional group; and (3) when tensions between the two can be resolved. I argue that transregional groups prefer to invest in local affiliates that are strong enough to provide secure access to territory, and have strong enough command and control over their sub-units to effectively carry out strategic operations. Strong local groups, however, don’t need transregional group support, and are unwilling to trade sovereignty and independence for any support the transregional group could offer. This creates a tension which is resolved when relatively strong local groups require goods that they cannot access themselves, but that transregional groups can provide. This occurs when strong local groups are in decline, or lose access to a resource critical to their survival. Understanding this tradeoff will help us to understand when local insurgencies are likely to become internationalized, and to exploit tensions in these partnerships.
The Bipolar Jihadist World Order: Exploring how competition between al Qaeda and IS impacts jihadist group alliance formation
Understanding why some groups have pledged allegiance to IS while others have either refrained or remained loyal to al Qaeda will help us to understanding the future of global jihadism. The emergence of the so called Islamic State (IS) has changed the very nature of global jihad not because it is particularly unique or because it brings new strategies to the table, but because of how its power and inevitable competition with al Qaeda has transformed the structure of the global jihadi system. This competition is inevitable because both groups are dependent upon their global reach in order to maintain their power. As conveyors of legitimacy and brokers of network goods, the strength of one necessarily undercuts the strength of the other.
I argue that before al Qaeda in Iraq (AQI) transformed into IS, the global jihadi system could be thought of as a hegemonic system dominated by al Qaeda. Al Qaeda, by nature of being the most prominent group, could selectively bestow legitimacy through signaling and formal membership within the al Qaeda franchise. Moreover, as the hub of a vast network, it had leverage over other groups in the system because it could choose to isolate others. The rise of IS can be conceptualized as the rise of peer challenger, transforming the system into a bipolar one. I argue that the structure of the system necessitates competition between IS and al Qaeda, and ask how this bipolarity impacts group decisions to enter the global system, and how it impacts their alliance decisions once they do.