sociogeNEsis of criMinal nEtworks: reconStruction, dIscovery and diSruption

Project Information

NEMESIS

Grant agreement ID: 101116718

DOI

10.3030/101116718

EC signature date 31 August 2023

Start date 1 January 2024

End date 31 December 2028

Funded under

• European Research Council (ERC)

Total cost

€ 1 498 794,00

EU contribution

€ 1 498 794,00

Coordinated by

UNIVERSITA COMMERCIALE LUIGI BOCCONI
Italy

Periodic Reporting for period 1 - NEMESIS (sociogeNEsis of criMinal nEtworks: reconStruction, dIscovery and diSruption)

Reporting period: 2024-01-01 to 2026-06-30

Summary of the context and overall objectives of the project

Criminal networks are different from any other networks studied in science. No other networks operate under the same objectives, efficiency-security trade-offs and adversarial nature that characterize criminal contexts. These distinctive aspects generate peculiar forms of incompleteness and structural complexity within the available criminal network data, which hinder progress in understanding how illicit organizations form, function, evolve and persist. While standard techniques from social network analysis have achieved important advances in this area, as leading scholars have noted, criminal networks are not simply social networks in criminal contexts.

NEMESIS embraces this challenge and transforms it into a unique opportunity to substantially advance the state of the art in criminal network analysis and modeling. Rather than viewing incomplete and partially-observed data as a limitation, the project interprets data missingness as a meaningful structural information that can reveal efficiency-security trade-offs, strategic redundancies, resilience mechanisms and obscured organizational boundaries. At the same time, it leverages the complexity inherent in observed criminal network data to infer higher-level topologies that uncover the sophisticated generative structures underlying organized crime. These include tree-based nested modular hierarchies among criminals, systematic architectures shared across multiplex interactions, and adaptive growth mechanisms. Collectively, these objectives aim to develop innovative analytical perspectives capable of formally capturing the mechanisms that shape the structure, evolution and resilience of criminal networks, ultimately supporting the design of more effective law-enforcement interventions.

To achieve the above objectives, NEMESIS integrates modern criminological and social science theories with rigorous statistical methodology and advanced computational techniques. The project designs innovative models for multiplex, multilayer and dynamic networks, combining evolutionary tree representations, zero-inflated processes, dependent partition structures and latent variable approaches. This integrated perspective expands current approaches to criminal network analysis and generates methodological innovations with relevance even beyond criminology. Its expected impact lies in advancing scientific understanding of organized crime and providing rigorous analytical foundations for more informed policy responses.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

To date, NEMESIS has developed an integrated framework that advances the state of the art in criminal network modeling, with particular focus on covert, incomplete, and complex connectivity patterns.

A first set of contributions addresses network reconstruction from incomplete data. Novel stochastic block models have been introduced to explicitly uncover efficiency-security trade-offs in weighted criminal networks, revealing how systematic absences of ties encode strategic redundancies and resilience mechanisms. These models were extended to multilayer and multiplex settings, enabling inference on structural determinants of group organization, prediction of unobserved ties, and analysis of how incoming affiliates integrate into criminal organizations. Complementary research will connect species-sampling processes from ecology with boundary specification problems to estimate the size of partially observed criminal organizations and infer the presence of unseen criminals. In parallel, algorithmic pipelines are being developed to automatically reconstruct multiplex criminal networks from judicial documents, addressing a major bottleneck in data availability and reproducibility.

A second line of work has leveraged the complexity inherent in observed criminal networks to uncover hidden higher-level architectures regulating criminal organizations. Phylogenetic latent space models were designed to integrate network representations with evolutionary tree structures, reconstructing hierarchical multiscale differentiation processes directly from observed criminal interactions. This framework reveals nested hierarchies, adaptive specialization patterns and long-term structural backbones, including those sustained by kinship and multiplex relations. Additional methodological advances ensure computationally efficient inference.

Building on these developments, the project has translated structural and evolutionary insights into analytical methods for identifying enablers and facilitators of crime, anticipating network adaptation and supporting disruption strategies that move beyond traditional centrality measures. Tensor-based models were also developed to analyze dynamic interdependencies among crime types across space and time, enabling network-aware hotspot analysis and more refined understanding of criminal activity patterns.

These achievements establish a coherent framework grounded in criminological theory and advanced statistical modeling.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

NEMESIS advances the state of the art in criminal network analysis in several directions.

First, it reconceptualizes missing data as structural information. Instead of treating partially-observed criminal networks as limited information, the project shows that systematic absences of ties can quantify efficiency–security trade-offs, strategic redundancies and resilience mechanisms. The proposed zero-inflated, multilayer and multiplex stochastic block models, combined with species sampling processes, make it possible to infer these hidden structures, quantify the size of the criminal organization and estimate the presence of unobserved criminals. This perspective redefines how incomplete criminal data are analyzed and provides a principled framework for studying covert organizations.

Second, the project introduces tree-based evolutionary representations of criminal networks. By integrating latent space models with phylogenetic trees, NEMESIS reconstructs modular multiscale differentiation processes that govern how criminals specialize, coordinate and adapt. This contribution represents a genuine transfer of ideas from phylogenetics to criminological network analysis, allowing inference on complex nested hierarchies among criminals and long-term structural backbones that remain invisible to traditional descriptive approaches. The relevance of this perspective extends beyond criminology.

The above developments advance the state of the art also in informing actionable policy-oriented strategies. Rather than relying on traditional centrality-based indicators, the project provides analytical frameworks that identify strategically positioned enablers, while accounting for redundancy and adaptation. This shift represents a substantial innovation in supporting disruption strategies. In addition, the newly developed tensor-based models capture dynamic interdependencies among crime types across space and time, enabling network-aware hotspot analysis that reflects cross-crime temporal spillovers. Together, these advances provide a more realistic and actionable understanding of complex criminal systems.

A further transformative contribution lies in the ongoing development of an NLP- and LLM-based pipeline for automated reconstruction of multiplex criminal networks from judicial documents. By addressing scalability and reproducibility, this contribution opens the possibility of large-scale analyses previously unfeasible in this field.