GenAISim: Simulation in the Loop for Multi-Stakeholder Interactions with Generative Agents

Project Summary

Traditional decision-making processes often struggle to adapt to the dynamic and multifaceted nature of the modern world. This research addresses a higher-level profound need for advanced automated decision-making tools that can address complex, context-rich challenges in society.

This project will investigate a hybrid decision-making system, leveraging cooperative knowledge from multiple stakeholders through socio-technical observations, and knowledge priors in Large Language Models (LLMs) and open datasets. It will develop GenAISim, a novel suite of generative and datadriven simulations, useful for depicting current and future urban scenarios, including in mobility, urban policymaking, and health domains. Through a multidisciplinary sociotechnical framework of investigation, this project will establish a new simulation in the loop paradigm.

Objectives:

  • Explore LLM agent-based synthetic data generation techniques to simulate and augment human behaviours in diverse contexts;
  • Develop a robust framework for hypothesis testing of real-world observations and relationships, while avoiding spurious correlations;
  • Investigate diverse stakeholder settings, often with nonoverlapping and potentially conflicting objectives, priorities, constraints, incentives and pain points; and
  • Explore questions around hybrid decision making – if an LLM agent is substituting for a decision maker in contexts.

Project team


Project partners & collaborators

 


Project funding

  • Australian Government through the Australian Research Council – ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S)


Abstract holographic human image standing in concentric circles.