LLMs in agent-based models, supply chain paradigms

Large Language Models in Agent-Based Simulations: Computer scientists evaluated the integration of large language models into agent-based models using the MultiVeStA statistical model checker. Modifying the Schelling segregation model in the Mesa framework showed that smaller models suffered operational failures under repeated tool-calling tasks, while larger language models maintained simulation stability.
Reductionist Paradigms in Supply Chain Management: An assessment of senior Australian supply chain practitioners applied a systems thinking construct termed the "spectrum of the whole." The study identified a structural divergence between practitioners' reductionist operational habits and the non-linear requirements of complex logistics systems.
Methodological Guidance for Transdisciplinary Research: A newly published scholarly volume provides protocols for cross-boundary scientific collaboration. The work details procedures at individual, interpersonal, and project scales, focusing on proposal development, conceptual integration, and research translation for complex societal problems.
Second-Order Cybernetics and AI Discourse: Researchers used second-order cybernetic frameworks to evaluate how large language models project neutrality while directing discourse. The analysis found that communicative responsibility is distributed through recursive feedback loops, with primary accountability resting on upstream actors who establish safety protocols and alignment parameters.
Evolution of Watershed Socio-Ecological Research: A PRISMA-guided review of watershed literature from 1999 to 2025 documented a post-2018 transition from conceptual outlines toward governance analysis, climate resilience, and spatial tools. The authors synthesized these dynamics into a unified framework connecting administrative processes with ecological thresholds.
Adaptive Cycle Heuristics in Subcultural Urban Spaces: Fieldwork conducted in Berlin's Kreuzberg district between 2023 and 2025 applied Holling's ecological adaptive cycle to urban geography. The study demonstrated how shifting measures of potential, connectedness, and resilience map trajectories in which symbolic value accumulation systematically exposes subcultural sites to institutionalization.
Self-Organized Criticality in Text Construction: Analysis of keystroke data from the KLiCKe corpus revealed that text autocorrelations adhere to power-law scaling with finite-size effects. Keystroke revisions produce size-dependent restoring dynamics that regulate semantic relationships toward a critical state during composition.
Stress-Redistribution in Coastal Ecosystems: Spatially explicit modeling of hummock-hollow meadow patterns showed that self-organization mitigates environmental pressure through stress redistribution. Precipitation carries salinity away from raised hummocks and into hollows, creating refugia that sustain higher plant biomass and extended phenological duration.
Frameworks for Adaptive Governance and Resilience: Recent literature introduced several targeted system frameworks: rural adaptive capacity was redefined as an interpretive and relational process; a coupled-systems model incorporated historical and socioeconomic variables into Caribbean coastal resilience; the IMPACT model was introduced for multi-agent historical simulations; road safety was mapped as a complex adaptive system; and the SOMA model was formulated for adaptive digital health governance.

Sources

Towards Agentic Agent-Based Models: Feasibility, Performance, and Statistical Model Checking
Abstract Agent-based models (ABMs) rely on simple, explicit and reproducible rules for individual decision making, while complex collective behavior emerges from interactions among agents. Recent advances in large language models (LLMs) make it tempting to replace, enrich, or perturb these rules with LLM-based agentic capabilities. However, this raises a methodological question: how does introducing LLM-driven decisions affect the reliability, computational cost, and behavior of ABM simulations? We investigate this for Mesa ABM models, a popular Python library for ABMs, analyzed by…
An exploration into reductionistic and holistic perspectives of supply chain practitioners using systems thinking: a mixed-methods study
Purpose Turbulent conditions confront supply chain practitioners, who respond by deploying interventions. Against this backdrop, is the paradigmatic perspective aligned with the supply chains they are part of? To what extent is there a misalignment between practitioner perspectives and the requirements of the system in which they operate? Design/methodology/approach Taking a systems thinking approach, this research examines 15 senior Australian supply chain practitioners through the application of a novel construct (spectrum of the whole), integrating knowledge from the transdisciplinary…
A Practical Guide to Transdisciplinary Research
Focusing on transdisciplinary research as a research method for the future, this volume provides the latest thinking and accessible resources for problem-solving, collaborative research that cuts across disciplinary boundaries and engages different communities. Transdisciplinary research is a problem-solving research approach where different disciplines are combined and transcended, primarily through teamwork, to solve complex problems in the real world innovatively. This volume teases apart what transdisciplinarity looks like in practice at the individual, interpersonal, and project levels.…
From cyberspeak to algorithmic newspeak: soft derogation and the ethics of AI-generated language
Purpose This paper examines the ethical and epistemological implications of language generated by large language models (LLMs) through the lenses of speech act theory, conversational implicature and second-order cybernetics. LLMs are often presented as neutral tools, yet their outputs actively participate in shaping discourse in ways that carry consequences for meaning, responsibility and pluralism. Design/methodology/approach The paper introduces the concept of soft derogation to describe a form of linguistic exclusion, one that operates through euphemization, omission and the simulation of…
Global Synthesis of Socio-Ecological System Research in Watersheds: A Systematic Review and Dynamic Framework Proposal
Watersheds are increasingly understood as socio-ecological systems (SES), reflecting the close interaction between ecological processes and human activities. However, research trends, thematic developments, and knowledge gaps in watershed-based SES studies have yet to be comprehensively synthesized. This study aims to examine the evolution of SES research in watershed contexts through a systematic review of the scientific literature published between 1999 and 2025. Following the PRISMA approach, records were retrieved from Scopus and Web of Science. A total of 689 documents were initially…
Subcultural space through the adaptive cycle – Resistance and commodification in Berlin Kreuzberg SO36
Urban transformation is commonly interpreted through processes of capital accumulation, gentrification, and displacement, while less attention has been paid to the temporal trajectories of subcultural spaces and their changing relationship with commodification. This study examines whether Holling’s adaptive cycle provides a useful analytical framework for interpreting these transformations in Berlin’s Kreuzberg SO36. The research combines literature and document analysis with fieldwork conducted during four visits between 2023 and 2025, including systematic site observation and photographic…
Writing as a Self-Organized Critical Process
We explain autocorrelation decay power laws omnipresent in texts by self-organized criticality. Specifically, we analyze the recently released KLiCKe keystroke dataset and show that not only the final texts' autocorrelations form a manifold that adheres to a power law with a finite-size scaling, but also the text revisions generate revision-size-dependent restoring dynamics toward that manifold. Thus, human writing appears to dynamically regulate semantic correlations in a text toward a critical state.
Self-organized hummock-hollow patterns redistribute salinity stress in coastal meadows
Self-organization theory predicts that ecosystem functioning and resilience can emerge from feedbacks that generate spatial patterns. These emergent properties have commonly been explained by local positive feedbacks that concentrate limiting resources within productive patches. Here, we identify a stress-redistribution feedback underlying self-organized hummock-hollow patterning. Through this feedback, salinity stress is partitioned across the patterned landscape, maintaining vegetated hummocks as low-salinity refugia alongside salt-rich hollows and thereby sustaining ecosystem functioning…

Also this week

Full transcript

Can agent-based simulations remain reliable once individual agents start making decisions through large language models? That methodological question leads our look at complex systems research today on ComplexityPod. We begin with computational modeling and statistical verification. Replacing rule-based agents in the Schelling segregation model with agents driven by language model tool calls brings a distinct failure mode to the surface. Because in Mesa simulations, agents execute sequential decisions. When smaller models attempted those sequential tool calls and classification steps, they broke down mid-run. Right. The larger models maintained operational stability throughout the iterations, but coupling the simulation with the MultiVeStA statistical model checker showed the trade-off immediately in the computational overhead. It turns the model checker into an automated verification gate. In agent-based modeling, macro-level patterns hinge entirely on micro-level consistency. If the language model fails its tool calls, the emergent segregation pattern tells you about software errors rather than collective behavior. That question of structural constraint shows up in accountability analysis as well. Looking at automated language systems through second-order cybernetics and speech act theory, language outputs operate within recursive socio-technical loops. And the analysis points out how those outputs project an appearance of neutrality while enacting what the authors term soft derogation. Through omission, euphemism, and simulated balance. Which redistributes responsibility. Even if users and deployers interact with the model day-to-day, the discursive boundaries are determined upstream by the teams building the alignment protocols and safety layers. Upstream constraints dictate the boundary conditions. We see boundary dynamics functioning in text production at the physical level, too. An analysis of keystroke records from the KLiCKe dataset found that text autocorrelation decay follows power-law scaling with finite-size effects. Operating essentially as a self-organized critical system. When a writer pauses and makes an edit, that revision acts as a restoring force proportional to the size of the modification. Pulling semantic correlations back toward the manifold, keeping the text structured right at that critical edge between order and disorder. Spatial persistence through localized regulation appears in physical ecosystems as well. Coastal hummock-hollow meadows use a stress-redistribution mechanism to survive salinization. Rainfall leaches salt out of the elevated hummocks and deposits it down into the hollows. So the hollows absorb the environmental stress, turning the hummocks into low-salinity refugia. In field measurements and remote sensing, those structured hummocks maintain higher biomass, extended phenological activity, and higher reproductive output than flat, homogeneous vegetation facing the same salt load. It is spatial heterogeneity absorbing systemic shocks. When you apply that kind of ecological model to urban systems, the trajectory shifts. Fieldwork conducted between 2023 and 2025 across three subcultural sites in Berlin's Kreuzberg district tracked spatial change using Holling's adaptive cycle framework. Tracking potential, connectedness, and resilience. But in Kreuzberg, accumulating symbolic value makes a space rigid. It increases its exposure to commercial redevelopment and institutional capture, driving the site into release and reorganization. That non-linear progression from resilience to capture mirrors broader socio-ecological monitoring. A review of 96 watershed articles published between 1999 and 2025 tracked how the field evolved from descriptive conceptual diagrams to quantitative frameworks. Especially after 2018, when studies began integrating geographic information systems with climate adaptation metrics and governance thresholds into dynamic watershed models. Yet that systemic understanding often fails to reach practitioners on the ground. A study of 15 senior Australian logistics managers revealed that most still rely on reductionist and functionalist paradigms. Even though supply networks are textbook complex adaptive systems. Operating with a linear mindset leaves organizations unprepared when systemic disruptions cascade through the distribution chain. Bridging that gap between systems models and practical action is the focus of a new methodological volume on transdisciplinary research. It sets operational protocols across individual, interpersonal, and project scales. Covering the full pipeline: joint research proposals, collaborative execution, and translating systemic outputs into community applications. Several other frameworks landed in this cycle targeting similar coupling problems. One reinterprets rural adaptive capacity as an ongoing, governed relational process, while another models Caribbean coastal resilience through combined historical, economic, and physical factors. We also saw IMPACT for multi-agent historical movement planning, a complex adaptive systems framework applied to road safety mapping, and SOMA, an adaptive governance framework for digital health. More research at the intersection of systems thinking and disciplinary science arrives next week. From ComplexityPod, thanks for listening.

From our community

At the Intersections — latest issue

Welcome to this week’s issue of At the Intersections. This edition connects investigations into biological and cognitive processes with observations of frontline community care. We look at the circulation of fluid within the brain and the n

Read the latest issue →