DISRUPTIVE METHODS AND TECHNOLOGIES

Different methods provide different lenses through which we can understand the world. Technological advancements enhance those lenses providing us with richer perspectives.

Modeling is at the center of the work we do from which we connect with approaches such as on-site observations, interviews, surveys, and simulation experimentation with technologies such as large language models and cloud computing.

Research Areas

Computational Modeling & AI

Uses data-driven and computational methods to study social behaviors and complex systems.

Analyzes patterns from large-scale digital, behavioral, and sensor data.
Models societal processes to generate insights for policy, governance, and intervention.

Agent and Simulation-Based Research

Simulates interactions of individuals or systems to explore emergent dynamics and outcomes.

Builds virtual environments where agents follow decision rules and interact.
Tests scenarios and interventions in controlled, replicable settings.

Large Language Models

Leverages advanced AI models to analyze, generate, and interpret human language data.

Applies LLMs for summarization, translation, sentiment, and policy analysis.
Integrates LLM outputs into larger simulation and decision-support systems.

Selected Publications

Enhancing GPT-3.5's Proficiency in Netlogo Through Few-Shot Prompting and Retrieval-Augmented Generation

Authors: Joseph Martínez, Brian Llinas, Jhon G Botello, Jose J Padilla, &Erika Frydenlund

Tags: Large Language Models, Simulation

The paper investigates the capabilities of GPT-3.5 in generating NetLogo agent-based modeling code for low-resource languages, using few-shot prompting and Retrieval-Augmented Generation (RAG) methods to enhance LLM performance.
GPT-3.5 showed improved NetLogo code generation when few-shot examples were provided, while the RAG approach performed poorly due to known limitations; the study outlines future research opportunities for integrating LLMs into agent-based model development and education.

Read paper

Modeler in a box: how can large language models aid in the simulation modeling process?

Authors: Erika Frydenlund, Joseph Martínez,

Jose J Padilla, Katherine Palacio, & David Shuttleworth

Tags: Agent-based Simulation, Large Language Models, Computational

The study explores whether ChatGPT can generate simulation model code (discrete event, system dynamics, and agent-based models) directly from prose-based narratives describing real-world scenarios, using a case study of transportation changes for elementary students during COVID-19.
ChatGPT struggled to produce concise or fully executable simulation models, especially in less familiar programming paradigms, highlighting both current limitations and the potential for future LLM improvements to support simulation modeling and interdisciplinary collaboration.