Dissertation · Collaborating Agents & Network Interdiction
How multi-agency collaboration improves labor trafficking interdiction outcomes
Overview
My dissertation develops network interdiction models to understand and improve how law enforcement agencies, NGOs, and government entities collaborate to disrupt labor trafficking in U.S. agricultural supply chains.
Advisor: Prof. Kayse Lee Maass, Operations Research & Social Justice Lab, Northeastern University
Committee: Dr. Shahin Shahrampour · Dr. Gabriela Gongora Svartzman
Expected completion: 2026
Motivation
Anti-trafficking organizations rarely operate in isolation; yet the operations research literature has largely modeled interdiction as a single-agent problem. In practice, Homeland Security investigators (DHS), labor departments (DOL), NGOs, and local law enforcement all play roles in detecting and disrupting labor trafficking, particularly in agriculture where H-2A guest worker programs create structural vulnerabilities.
What happens when these agencies are acting in the same trafficking operations? How does the effect of their interdictions change because they are all unintentionally interfering with each other’s actions? For example, how can we measure the effect of disruption on trafficking when the DHS deport the immigrant laborers who were serving as witnesses in an FBI or DOL case? These are the central questions of my dissertation.
The four papers
Paper 1, the scoping review. Multiple non-independent interventions on the same network element. A PRISMA-ScR review registered on OSF, across four databases with two independent reviewers. It establishes that independence, or the absence of same-element multiplicity altogether, is the norm in the literature, which is the gap the rest of the dissertation addresses. The registered protocol is available as a PDF.
Working paper with K. L. Maass. Revision complete, submitting to Networks, September 2026
Paper 2, the model. [Job market paper] Collaboration in anti-trafficking efforts: a network interdiction problem with multiple dependent attacks. Rather than committing to one algebraic form for how interventions combine, it defines the class of effect functions satisfying four axioms and shows the standard independence assumption sits at the boundary of that class. Two results are aimed at a management audience: the cost of assuming independence, and the coordination gap between a centralized task force and the same organizations spending the same total separately.
Working paper with K. L. Maass. Results analysis. Target: Production and Operations Management
This paper is the work partially funded by the DHS CINA project, Network Interdiction Models to Improve Effectiveness of Coordinated Labor Trafficking Disruption Efforts, PI Kayse Maass
Paper 3, theory and complexity. Optimal bounds and complexity for network interdiction with dependent multiple interventions. Under a logarithmic transform the problem is shortest-path interdiction and therefore NP-hard. We show that depending on the structure, the problem is polynomial and independent of the choice of function, and that the optimal intervention can be restricted to a candidate set determined by which parts of the network are in play.
Working paper with K. L. Maass. Results analysis. Target: IISE Transactions
Paper 4, the algorithm and computational study. Algorithms for network interdiction problems with non-convex functional attacks. A three-phase procedure, structure recovery, depth determination, and budget allocation with re-optimization after each commitment, verified against exhaustive enumeration and compared with a baseline heuristic. Working paper with K. L. Maass. Model and algorithm ready, coding in progress
Related work
The empirical companion. Network interdiction to improve labor trafficking detection in the U.S. agricultural sector, grounded in the H-2A violation data modeled in our PLOS ONE paper (Jafari et al., 2024).
Revise and resubmit at Decision Sciences. With Jafari, Bhimani, Farrell and Maass
Picking up after the degree. Unintended consequences of enforcement in labor trafficking networks: geographic displacement, operational fragmentation, and compliance switching. How enforcement that looks optimal for one agency can displace an operation or fragment it rather than disrupt it.
Research in progress with K. L. Maass. Ideation and literature review, resuming after degree completion
An auto-ethnographic account of a student’s experience as a teaching fellow in industrial engineering An auto-ethnographic research of my experience in the Teaching Fellowship program at Northeastern University and how it affected my development as a teacher. Independent research in progress. Data collection, cleaning, and initial literature review ready; resuming after degree completion.
Methods
- Network interdiction (bilevel programming, mixed-integer programming)
- Multi-agent optimization (cooperative and non-cooperative settings)
- Python with Gurobi for computational experiments
- Public datasets: H-2A violation records, DOL enforcement data (500,000+ records)
Selected Presentations
- INFORMS Annual Meeting 2025 — Atlanta, GA (podium talk)
- INFORMS Annual Meeting 2024 — Seattle, WA
- INFORMS Annual Meeting 2023 - Phoenix, AZ
- INFORMS Annual Meeting 2022 - Indianapolis, IN
- RISE Research Expo, Northeastern University 2024 (poster)
- MIE Research Expo, Northeastern University 2023 (poster)