Incentives to Build Houses, Trade Houses, or Trade House Building Skills in Simulated Worlds under Various Governing Systems or Institutions: Comparing Multi-agent Reinforcement Learning to Generative Agent-based Model

TL;DR

This paper simulates two somewhat similar worlds using multi-agent reinforcement learning (MARL) framework of the AI-Economist and generative agent-based model (GABM) framework of the Concordia to simulate a similar social phenomena with limitations.

Abstract

It has been shown that social institutions impact human motivations to produce different behaviours, such as amount of working or specialisation in labor. With advancement in artificial intelligence (AI), specifically large language models (LLMs), now it is possible to perform in-silico simulations to test various hypotheses around this topic. Here, I simulate two somewhat similar worlds using multi-agent reinforcement learning (MARL) framework of the AI-Economist and generative agent-based model (GABM) framework of the Concordia. In the extended versions of the AI-Economist and Concordia, the agents are able to build houses, trade houses, and trade house building skill. Moreover, along the individualistic-collectivists axis, there are a set of three governing systems: Full-Libertarian, Semi-Libertarian/Utilitarian, and Full-Utilitarian. Additionally, in the extended AI-Economist, the Semi-Libertarian/Utilitarian system is further divided to a set of three governing institutions along the discriminative axis: Inclusive, Arbitrary, and Extractive. Building on these, I am able to show that among governing systems and institutions of the extended AI-Economist, under the Semi-Libertarian/Utilitarian and Inclusive government, the ratios of building houses to trading houses and trading house building skill are higher than the rest. Furthermore, I am able to show that in the extended Concordia when the central government care about equality in the society, the Full-Utilitarian system generates agents building more houses and trading more house building skill. In contrast, these economic activities are higher under the Full-Libertarian system when the central government cares about productivity in the society. Overall, the focus of this paper is to compare and contrast two advanced techniques of AI, MARL and GABM, to simulate a similar social phenomena with limitations.

Figures (31)

• Figure 1: The central social planner of the AI-Economist is extended in two directions. First, along individualistic-collectivistic axis, meaning that how much it respects the decision power of the mobile agents in the society, it is divided to three governing systems: Full-Libertarian, Semi-Libertarian/Utilitarian, and Full-Utilitarian. Second, along discriminative axis, meaning that how much it considers the voting of all mobile agents in the society equally, it is divided to three governing institutions: Inclusive, Arbitrary, and Extractive.
• Figure 2: A schematic figure showing the environment of the extended AI-Economist used in this paper. In all simulations of this paper, there are six agents in the environment which simultaneously cooperate and compete to gather and trade three natural resources (wood, stone, and iron), using them to build three types of houses (red, blue, and green), trade these three types of houses, and trade house building skill to earn incomes. Moreover, they rank the three material resources. Also, at the end of each tax period, they pay their due taxes to the central social planner. The central social planner optimises its own reward function which could be a combination of equality and productivity in the society, and depending on the governing system or institution, it counts the votes of the agents and invests the total collected due taxes on the three material resources accordingly.
• Figure 3: Various initialisation parameters of the extended Concordia.
• Figure 4: The ratio of building houses to trading houses (left) and the ratio of building house to trading house building skill (right) across three governing systems of the extended AI-Economist: Full-Libertarian, Semi-Libertarian/Utilitarian, and Full-Utilitarian. The ratios are obtained by averaging over two similar simulations per governing system differing only in the type of the reward function of the central planner (Fig. \ref{['Figure15']}). As it is clear from this plot, in both left and right panels, the ratios are higher for the Semi-Libertarian/Utilitarian governing system.
• Figure 5: The ratio of building houses to trading houses (left) and the ratio of building house to trading house building skill (right) across three governing institutions of the Semi-Libertarian/Utilitarian governing system of the extended AI-Economist: Inclusive, Arbitrary, and Extractive. The ratios are obtained by averaging over two similar simulations per governing institution differing only in the type of the reward function of the central planner (Fig. \ref{['Figure15']}). As it is clear from this plot, in both left and right panels, the ratios are higher for the Inclusive governing institution.