Large Language Models for Agent-Based Modelling: Current and possible uses across the modelling cycle

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: We're looking at the paper: "Large Language Models for Agent-Based Modelling".

Jane: Here's the gist of the paper. The gist: This study reflects on the opportunities and challenges of Large Language Models (LLMs) in Agent-Based Modelling (ABM) by following the modelling cycle from problem formulation to documentation and communication of model results, holding a critical stance.

Tom: Conceptual Background Agent-based Modelling (ABM) provides a bottom-up framework for simulating complex social phenomena through the behaviors and interactions of individual agents in a given environment Social simulation is a methodology for investigating complex social phenomena through the construction and analysis of artificial societies A common approach used within social simulation is Agent-based Modelling (ABM), which involves modelling individual entities (agents) that represent social actors such as individuals, groups, or organizations ABM usually emphasizes bottom-up dynamics whereby macro-level patterns emerge from micro-level interactions of heterogeneous agents ABM is a multi-phase process involving abstracting from the world / phenomenon to the target system, understanding this system, abstracting concepts, modeling relationships, and creating simulation outputs A generic ABM cycle is presented in Figure 1 Each phase consists of several iterative steps and sub-steps This iterative process applies to the main phases (refinement through multiple cycles) and also within each phase by continuous improvement in feedback loops Large Language Models (LLMs) are advanced neural networks pre-trained on vast textual datasets and used to process and generate human-like language They typically fall into two categories: generative models (e.g., GPT) and non-generative models (e.g., BERT) Generative models predict the next token in a sequence, enabling them to produce coherent, contextually appropriate text In contrast, non-generative models are commonly used for quantifying text into mathematical vectors These vectors have the property of being geometrically close when the original texts from which they are derived are semantically close, making them well-suited for tasks like classification and retrieval of similar information Although both capture deep linguistic patterns, only generative LLMs are designed for open-ended text generation and dialogue There are several main usage areas for LLMs

48: , including natural language understanding (LLMs exhibit strong generalization capabilities), text and code generation (LLMs can produce coherent, contextually appropriate, and high-quality text and code, supporting applications such as content creation, programming assistance, and automated writing), knowledge-intensive tasks (LLMs can encode extensive and possibly implicit domain-specific knowledge, enabling them to perform tasks that require broad world knowledge or specialized expertise), translations and transcriptions, and reasoning and problem solving (LLMs can support complex tasks through pattern-based inference, analogical reasoning, and step-by-step problem decomposition) The use of LLMs has been associated with several ethical concerns

23: , including bias and fairness, privacy and data security, misinformation and disinformation, transparency and accountability, intellectual property and plagiarism, access and inequality How it works The study addresses two research questions: 1.

Jane: Where and how are LLMs used in the ABM cycle?

Paper summary: Tom: So we're looking at this paper titled "Large Language Models for Agent-Based Modelling: Current and possible uses across the modelling cycle." Basically, the authors are pointing out that there's no clear structure for how to use Large Language Models with Agent-Based Modelling. They want to explore where these two things actually fit in the whole process of building an ABM model <ref:2507.05723#pg2>.

Jane: That's right, and their main point is that they're going to follow the entire modelling cycle—from how you start a problem all the way to documenting and communicating your results—to see where LLMs can actually help at each stage <ref:2507.05723#pg3>.

Lu: They claim that by using the structure of the ABM cycle instead of just following whatever structure an existing paper uses, they can look at all the phases and consider how different LLM abilities, like summarizing or reasoning, might apply to each one <ref:2507.05723#pg3>.

Meng: So what's actually interesting is that they identify that most of the current use in ABM papers focuses on implementation, specifically using agents powered by LLMs for things like reasoning or decision-making <ref:2507.05723#pg2>.

Lalam: They also found one study where LLMs were used to generate narrative explanations to communicate what agents are thinking about key life events in a model <ref:2507.05723#pg1>.

Tom: So the thesis is that we need a structured approach because right now, the potential of LLMs across the whole cycle just isn't being fully leveraged <ref:2507.05723#pg1>.

Jane: It really matters because it shows that we can't just throw an LLM into an ABM project and expect magic; you have to consider how it fits in every single step of building the simulation <ref:2507.05723#pg3>.

Conclusion: Tom: Looking at "Large Language Models for Agent-Based Modelling: Current and possible uses across the modelling cycle," it seems like the core message is that LLMs are a powerful tool, but we need to be deliberate about how we use them in the modeling process <ref:2507.05723#pg1>.

Jane: The authors are suggesting that by systematically looking at each phase of the ABM cycle, you can figure out exactly where an LLM’s strengths—like generating text or performing reasoning—will give you the most help <ref:2507.05723#pg3>.

Lu: They are essentially arguing that we need to move past just seeing one application and look at the whole pipeline, from formulating the problem to communicating what we found <ref:2507.05723#pg1>.

Meng: For someone trying to actually build something, it implies that you don't just need an agent with a smart brain; you need a plan for how that brain interacts with the rest of your simulation structure <ref:2507.05723#pg3>.

Lalam: It suggests that the real utility isn't just in the agents themselves, but in using LLMs to help handle those messy parts like literature reviews or translating stakeholder notes into model rules <ref:2507.05723#pg1>.

Tom: So, the implication is that if we want to get more useful agent-based models out there, we need to stop treating LLMs as a tacked-on feature and start integrating them into the entire modeling workflow <ref:2507.05723#pg1>.