Generative Social Science seeks to explain complex socioeconomic phenomena by simulating micro-level agent behavior and observing emergent macro-level patterns. Inverse Generative Social Science (iGSS) reverses this logic: beginning from empirically observed or theoretically established macro-level outcomes, it uses computational optimization to search for the micro-level rules that could have produced them. While iGSS has so far been applied mostly to narrowly defined empirical problems, we argue that it holds broader promise for foundational questions of social and economic theory — including the role of rules and equilibria in institution formation, the microfoundations problem in economics, and theories of social roles and self-organization.

We illustrate this potential by using iGSS to reverse-engineer a recent agent-based model of the evolution of human cooperation, in which well-studied direct and indirect reciprocity mechanisms are contrasted to the institution of money-mediated exchange. Rather than implementing predetermined decision rules, we endow agents with memory, reputations, and tokens, and enable them to discover possible action rules through genetic programming, investigating to what extent these discovered rules correspond to mechanisms established in the cooperation and monetary theory literatures.

Preliminary qualitative results across three testing modes are encouraging. When endowed with signals and assessment mechanics, iGSS agents discover action rules corresponding to Indirect Reciprocity and Money mechanisms reliably, and less so for Direct Reciprocity; when all signals are available simultaneously, both monetary and reputational strategies appear, as well as hybrid combinations. When only the IR action rule is fixed, iGSS can discover assessment rules coherent with well-studied Standing norms, with results suggestive of leading eight IR strategies in some runs. When provided only with an agnostic standing signal, iGSS agents can in some cases co-evolve both action and assessment rules recognizable as indirect reciprocity mechanisms, though this remains less reliable and appears sensitive to ecological conditions.

We close by reflecting on the deeper challenge posed by money. Unlike reciprocity mechanisms, monetary exchange requires agents with genuine social roles and interdependent decision rules — a form of coordinated institutional action that exceeds what standard representative-agent iGSS architectures can currently capture. Solving this limitation, we argue, is precisely what would make iGSS a genuinely powerful tool for theoretical social science.

This contribution describes and assesses fundamental quantum computing concepts from the perspective of the social sciences. The property of qubits to exist in superposition makes quantum computers particularly suited for simulating systems with a high amount of uncertainty. In addition, the property of superposition makes the evaluations of multiple configuration at the same time possible, increasing the processing speed of quantum computers. We argue that

these properties make quantum computers a potentially useful tool for social science modellers.

Therefore, in this contribution, we introduce how to translate an opinion dynamics (OD) model into quantum computer logic and demonstrate this translation process with an example model. We show that the model replicates previous results in OD research, in that higher empathy leads to faster opinion change, while also offering new insights, in that the process rather than a converged final state shows the main effect.

It is widely thought that LLM-based code generation can increase the productivity of researchers by helping them focus on providing

innovative solutions to important problems rather than having to deal

with syntax and implementation details. In order to study the plausibility and effectiveness of such a scenario for social simulation, we develop

an online prototype of an agent-based simulation environment that integrates Generative-AI (GenAI) in the creation of code from structured

natural language (NL) descriptions of models. Our system uses these

descriptions to automatically synthesize prompts and guide the LLM to

generate code in our special purpose modeling language. In order to solve

common problems related to LLM-based code generation, we have developed an interpreter that automatically analyzes the generated code to

verify its syntactic correctness and examine its correspondence with the

NL descriptions. If the interpreter detects no errors, the system automatically creates a detailed description of the code through which the user

can also verify its correctness. In addition, it provides an environment for

the safe execution of the model code. Our interaction with this prototype

indicates that LLM-based automation, when performed in a structured

and tightly controlled simulation environment, is plausible and can relieve the researcher from programming and troubleshooting simulation code.

Large Language Models are increasingly used to create generative agents for behavioral simulations and agent-based models, as they allow researchers to build entities that exhibit seemingly human-like behavior with relatively low effort. Although generative agents are frequently parameterized using empirical data, systematic empirical validation remains scarce: there is no established practice of benchmarking agent behavior against human reference data, nor a common standard for reporting validation in a way that allows comparison across studies. No dedicated infrastructure exists to support the collection, documentation, and reuse of validated agents. As a result, generative agents are difficult to transfer across studies, hard to reproduce, and lack comparable validation – ultimately limiting cumulative scientific progress. In response, we present ValidAgent, a prototype of an open, web-based repository that provides a common interface for researchers to browse, compare, and select from a curated collection of generative agent profiles. The platform emphasizes empirical validation, transparency, and reproducibility by requiring each agent profile to include structured documentation covering its design rationale, behavioral traits, underlying assumptions, intended scope of application, and validation results against human reference data. To demonstrate the platform’s utility, we provide an initial set of empirically validated agents grounded in the die-roll honesty paradigm, a widely used experimental setup for studying dishonest behavior. Building on this, we aim to foster a researcher-driven ecosystem in which generative agents can be contributed, peer-reviewed, and benchmarked – ultimately establishing citable generative agent sets. In doing so, this initiative contributes to a more transparent and cumulative social simulation practice.

This work explores the use of behaviour trees as a framework for modelling agent behaviour in agent-based models. While widely used in robotics and game development, behaviour trees remain underutilised in social simulation. We provide a brief introduction to the formalism and illustrate its application through a fisheries ABM of purse-seine vessels operating along the Catalan coast. The example demonstrates how behaviour trees can represent sequential decision processes, support modular design, and facilitate the reuse of behavioural components. We argue that behaviour trees offer a flexible and scalable alternative to more traditional approaches, with potential to contribute to ongoing efforts toward reusable building blocks in ABM.