From Heard to Lived Opinions: Simulating Opinion Dynamics with Grounded LLM Agents in Economic Environments

Abstract

Opinion dynamics (OD) studies how individual opinions evolve and generate collective patterns such as consensus and polarization. While recent work explores OD using populations of LLM-based agents focusing on opinion exchange, it typically does not incorporate individuals' lived experiences, such as economic outcomes of past decisions, which play a critical role in shaping opinions. We propose a novel OD simulation framework that grounds LLM-based agents in an economic environment, allowing them to act and receive environmental feedback. Our simulations exhibit coherent OD at both individual and population levels: individual opinions follow structured trajectories shaped by economic experiences, with adverse conditions inducing opinion rigidity, while at the population level, collective opinions co-move with economic conditions, with inequality amplifying polarization and price instability driving larger distributional shifts. These results highlight the importance of grounding LLM-based agents in environments to capture collective OD.

Figures (5)

• Figure 1: Structure of the LLM-based OD simulations. At each time step $t\in\{1,...,T\}$, LLM-based household $i\in\{1,...,n\}$ sequentially makes economic decisions regarding labor supply $L_{t,i}$ and consumption $C_{t,i}$. A rule-based firm produces goods $Y_{t,i}$ using household's labor and sets wage $\bar{w}_{t,i}$ and price $\bar{p}_{t,i}$ according to demand-supply imbalance. A government levies taxes $\tau_{t,i}$ and provides subsidies $s_{t,i}$ to households. In parallel, LLM-based agents exchange their opinions $O_{t,i}$ about the societal and economic environment.
• Figure 2: Household labor supply and consumption choices across economic conditions. Each panel reports the distribution of categorical actions ($a^L_{t,i}$ and $a^C_{t,i}$) for labor and consumption.
• Figure 3: TF-IDF-based word clouds of household opinion texts across four internal states.
• Figure 4: Representative internal state transition patterns obtained by clustering transition count vector $\bm{n}_i$ across all simulations. Each panel visualizes the representative agent’s state transition matrix as a heatmap, where rows (From) indicate the internal state $h_{t-1,i}$, columns (To) indicate $h_{t,i}$, and color intensity denotes transition frequency.
• Figure 5: Relationship between different types of opinion change rates $\delta_t,\delta_t^{'}$ and price fluctuations. Absolute price return means the corresponding average absolute log price change $|\log p_{t,i}/p_{t-1,i}|$. Blue circles count all sentiment changes $\delta_t$, whereas orange squares exclude transitions to the neutral (NEU) class $\delta_t^{'}$. Only bins with more than 100 observations are shown. Error bars indicate the standard error of the mean (SEM).