AI in Complexity Economic Modeling

Complexity economic modeling (CEM), particularly agent-based modeling (ABM), emerged as a powerful new tool to simulate and understand the behavior of large-scale economic systems. CEM models rely on the interactions of many heterogeneous agents, each with distinct characteristics and decision-making processes. As models scale to include millions of agents, the complexity grows exponentially, making it challenging to determine the optimal set of attributes and behaviors to model. This is where modern large language models (LLMs) can play a transformative role.

LLMs, trained on vast amounts of data, can assist in the conceptualization, design, and refinement of CE models by suggesting the characteristics and factors that should be incorporated into the agents. By analyzing existing literature, datasets, and economic theories, LLMs give insight into which variables and relationships might be most impactful for specific economic phenomena. Here, we propose a framework for leveraging LLMs to suggest characteristics for a large-scale (1 million agent) economic model.

Complexity Economics

Complexity economics (CE) represents a shift from the linear, equilibrium-focused approaches of classical economics to a dynamic and interconnected view of economic systems. Unlike classical economics, which often assumes rational agents operating in a static environment, complexity economics acknowledges that real-world economies are composed of diverse agents interacting in complex, adaptive ways. These interactions lead to emergent phenomena such as bubbles, crashes, and network effects that are difficult to predict using traditional models.

Comparison with Classical Economics:

• Agent Behavior: Classical economics typically assumes rational agents with perfect information, making optimal decisions. Complexity economics, however, models agents as boundedly rational, often relying on heuristics and subject to behavioral biases.
• Equilibrium vs. Dynamics: Classical models often focus on equilibrium states, where supply meets demand. Complexity economics, in contrast, is more concerned with the dynamics of the system, exploring how economies evolve over time and how new patterns emerge from the interactions of agents.
• Top-Down vs. Bottom-Up: Classical economics often uses top-down models where aggregate outcomes are derived from a set of predefined rules. Complexity economics takes a bottom-up approach, simulating the interactions of individual agents to observe how macroeconomic phenomena emerge from micro-level behaviors.

Current Applications: Complexity economics has been increasingly applied in various domains, including financial markets, urban development, and environmental sustainability. During the COVID-19 pandemic, complexity economics played a vital role in modeling the spread of the virus and its economic impacts. Agent-based models helped policymakers understand how different intervention strategies could affect not just public health outcomes but also economic resilience. These models provided insights into how local interactions, such as social distancing and vaccination, could scale up to influence broader economic stability.

Reception Among Classical Economists: Classical economists have been both intrigued and critical of complexity economics. Some praise its ability to capture the nuances of real-world economic behavior, particularly in situations where classical models fall short. However, others caution that complexity economics can become overly descriptive and lack the predictive precision that traditional models offer. Despite these debates, complexity economics is gaining traction as a complementary approach, particularly in understanding economic phenomena that are difficult to explain through classical lenses alone.

Proposed Framework for LLM-Assisted Characteristic Selection in Agent-Based Models

1. Literature Review and Knowledge Extraction

• Objective: Identify core and emerging factors in economic modeling.
• Process: LLMs can rapidly analyze a vast corpus of academic papers, books, and reports on economic modeling and related disciplines. By summarizing key findings and trends, LLMs can highlight the most commonly used variables, as well as new, underexplored factors that may be relevant to the model.
• Output: A comprehensive list of potential characteristics, categorized by their frequency of use and relevance to the economic phenomena being modeled.

2. Contextual Relevance Analysis

• Objective: Tailor characteristics to the specific economic context.
• Process: By providing the LLM with details about the economic environment, such as the type of market, geographic region, or historical context, the model can suggest particularly relevant characteristics. For instance, in a model focused on developing economies, factors such as informal labor markets or microfinance availability might be crucial.
• Output: A refined list of characteristics that are contextually relevant to the specific economic scenario being modeled.

3. Agent Heterogeneity Design

• Objective: Design a diverse set of agent types.
• Process: LLMs can help design the heterogeneity of agents by suggesting variations in demographics, preferences, risk aversion, access to information, and decision-making strategies. By considering these differences, the model can better capture the diversity of behaviors observed in real-world economic systems.
• Output: A detailed specification of different agent types, including their key characteristics and behavioral rules.

4. Behavioral and Interaction Dynamics

• Objective: Define the rules governing agent behavior and interactions.
• Process: LLMs can assist in defining the decision-making processes of agents, including how they interact with each other and respond to environmental changes. This might involve suggesting specific algorithms or heuristics that agents use to make decisions, based on psychological and sociological insights.
• Output: A set of behavioral rules and interaction dynamics that govern the actions of agents within the model.

5. Scenario Testing and Sensitivity Analysis

• Objective: Evaluate the impact of different characteristics on model outcomes.
• Process: LLMs can help design experiments to test how different combinations of characteristics affect the overall behavior of the economic system. By running simulations with varying agent characteristics, the modeler can assess the sensitivity of the results to changes in key factors.
• Output: A set of recommended characteristics that have the most significant impact on model outcomes, along with insights into the robustness of the model under different scenarios.

Conclusion

Incorporating LLMs into the process of developing complexity economic models offers enhanced sophistication and accuracy of agent-based models. By systematically suggesting characteristics and behaviors to model, LLMs help researchers and policymakers build more nuanced and effective simulations of economic systems. This framework provides a structured approach to leveraging LLMs for the selection of agent characteristics, ultimately leading to more robust and insightful economic models. As LLMs continue to evolve, their role in economic modeling is likely to expand, offering even greater potential for innovation and discovery in the field.

?