Research, Publications, and Academic Work

The research lineage begins with an undergraduate honors thesis at Simon's Rock College of Bard: Attentional Intensity in the Auditory Neurosystem, Spectral Space: Background, Current Theory, and New Research. The thesis received Institutional Review Board approval for human subjects testing, involved two years of experimental work on auditory attention and cognitive decision-making, and required the development of dedicated software for three interrelated experimental trials. It was an early investigation of the question that would animate the next two decades of work: how do complex systems, biological and social, produce behavior that cannot be predicted from their components alone?

Graduate work at the University of Connecticut extended these questions to organizational systems. The master's thesis applied network science and graph theory to the problem of organizational resilience: how do institutions fail under stress, and what structural conditions make them more robust? The thesis was sponsored by the Department of Defense through the Homeland Security Leadership program and is maintained as part of the Homeland Security Digital Library. This work is the direct intellectual precursor to the AI iatrogenics and robustness gap frameworks that now anchor the 1023AI practice.

The NPS period produced the CASS framework in its original form and the bulk of the published research record. As a Principal Investigator and DoD civilian program leader in the MOVES Institute, the work addressed a genuine operational need: defense and policy decision-makers required tools that could model how civilian populations in conflict environments respond to military operations, political change, and information campaigns. No existing simulation framework adequately captured the emergent dynamics of these systems.

CASS addressed this by treating societies as complex adaptive systems: modeling the beliefs, values, and interests of population segments as agent-level state variables, building social network structures from empirical survey data, and simulating how information, events, and interventions propagate through those networks to produce collective behavioral outcomes. The framework was applied to counterterrorism, counterinsurgency, peacekeeping, and stability operations contexts, with validation conducted against real-world outcomes in multiple international environments.

The research was presented at major international venues including the Winter Simulation Conference, the AAAI Spring Symposium, the Spring Simulation Multi-Conference, and the International Network for Social Network Analysis Sunbelt conferences in Italy and the United States. Publications appeared in peer-reviewed journals including Simulation and the Journal of Defense Modeling and Simulation, and in Springer's Lecture Notes in Computer Science series.

The transition from NPS to Northrop Grumman was not a departure from the research program. It was its largest-scale application. As Senior Manager and Senior Analyst supporting the Defense Manpower Data Center, the same complex adaptive systems principles developed for the CASS framework were applied directly to the design and architecture of enterprise-scale federal systems operating within a funding environment exceeding $60 billion.

Two programs stand out as direct applications of complexity science at institutional scale. The Global Force Management (GFM) system and the GFM Data Initiative (GFM-DI) are DoD enterprise systems responsible for tracking, allocating, and managing the full scope of US military forces globally, a genuinely complex adaptive system with thousands of interdependent entities, nonlinear interaction dynamics, and emergent force posture outcomes that no single component produces alone. The architectural decisions made during this period, including the transition from legacy relational structures toward flat-file and document-based storage systems such as MongoDB, were grounded in an understanding of how complex data relationships behave at scale and under stress, specifically, the brittleness of tightly coupled relational schemas when the system they model does not itself behave relationally.

The Department of Defense and Department of Veterans Affairs Electronic Health Record modernization programs presented a parallel set of complexity challenges at the intersection of technical architecture and human systems. EHR systems are not software problems. They are sociotechnical systems problems: the technical architecture determines what clinicians can see and do, the organizational dynamics around the system determine whether what they see is accurate and timely, and the compliance and security infrastructure determines whether the whole system holds under the adversarial pressure federal health data systems face continuously. The consistent thread across all programs was a design philosophy grounded in complex systems thinking: favor modularity over tight coupling, build for the unexpected interaction rather than the anticipated one, and treat the human system operating around the technical system as a core architectural variable.

See the full treatment at Live Market Operations →. In brief: the CASS methodology was extended to live US equity index derivatives markets as an empirical validation environment for the core theoretical claims. Three years of simulation-only operation preceded live capital deployment. The system operated across ES and NQ futures and futures options from 2015 to 2023, producing validated performance results across multiple market regimes including the March 2020 COVID crash.

The connection to the broader research program is methodological, not analogical. The same agent-based framework that modeled emergent behavior in conflict dynamics was applied to emergent behavior in financial markets. The same pre-deployment simulation discipline that the AI safety community advocates for capable AI systems was applied before any live capital was risked. The results validated the central thesis: that complex adaptive systems producing emergent behavior under stress are modelable in advance, governable through disciplined simulation, and fundamentally different from what static testing frameworks can detect.

The current research program applies complex adaptive systems methodology directly to the governance of capable AI. Two parallel tracks are active simultaneously: doctoral research at USC's Suzanne Dworak-Peck School of Social Work, and independent theoretical work on AI alignment fragility and governance architecture through 1023AI.

The doctoral track centers on a capstone examining LLM governance in professional social work practice environments. The core argument: the decisive barrier to safe AI integration in high-stakes professional contexts is a governance architecture problem, not a technical one. The intervention point is not better models or better prompts. It is the design of the institutional and social structures that determine whether safety frameworks hold under real operational pressure. The capstone applies antifragile systems theory, drawing centrally on Taleb's fragility framework, alongside established community-based participatory research and co-design methodology and the current AI safety literature, to argue for a specific governance architecture: antifragile, community-based participatory co-design as the necessary mechanism for safe LLM deployment in professional practice. Expected completion: 2027.

The theoretical track is described in the Theoretical Contributions section below.