This paper studies the generation of inflation expectations using generative AI in survey experiments, specifically focusing on agents created with Large Language Models (LLMs). It reveals that LLMs accessing relevant economic documents exhibit more consistent responses, demonstrating that domain knowledge significantly shapes expectations. Typically, LLMs predict higher inflation than actual rates, aligning closely with findings from the Survey of Consumer Expectations. Information treatments, particularly forward guidance, precisely influence LLMs' inflation expectations. Furthermore, customizing prompts with a persona induces partisan biases, mirroring human survey behaviors. It also explores the importance of model selection and the reasoning behind LLM responses, emphasizing how these elements impact outcomes. Additionally, the paper shows that LLMs can generate granular predictions of inflation expectations across demographic groups, suggesting potential applications in forecasting survey responses with limited data.

This study examines risk and time preferences of large language models (LLMs) in economic decision-making scenarios by adapting established behavioral economics measures. Using a novel methodology combining predefined risk personas and randomized controlled trials (RCTs), we assess AI agents' responses to various economic contexts. We find that AI models demonstrate systematic preference patterns influenced by both assigned personas and economic conditions, with larger models exhibiting more sophisticated preference updating than smaller models. Notably, recession scenarios elicit stronger preference adjustments than boom scenarios across all models, while risk-averse personas maintain consistent positions regardless of context. The results reveal that AI agents can develop coherent economic preferences distinct from human decision-making patterns, particularly in their response to negative economic information.

We simulate economic forecasts of professional forecasters using large language models (LLMs). We construct synthetic forecaster personas using a unique hand-gathered dataset of participant characteristics from the Survey of Professional Forecasters. These personas are then provided with real-time macroeconomic data to generate simulated responses to the SPF survey. Our results show that LLM-generated predictions are similar to human forecasts, but often achieve superior accuracy, particularly at medium- and long-term horizons. We argue that this advantage arises from LLMs' ability to extract latent information encoded in past human forecasts while avoiding systematic biases and noise. Our framework offers a cost-effective, high-frequency alternative that complements traditional survey methods by leveraging both human expertise and AI precision.

We directly compare the persuasion capabilities of a frontier large language model (LLM; Claude Sonnet 3.5) against incentivized human persuaders in an interactive, real-time conversational quiz setting. In this preregistered, large-scale incentivized experiment, participants (quiz takers) completed an online quiz where persuaders (either humans or LLMs) attempted to persuade quiz takers toward correct or incorrect answers. We find that LLM persuaders achieved significantly higher compliance with their directional persuasion attempts than incentivized human persuaders, demonstrating superior persuasive capabilities in both truthful (toward correct answers) and deceptive (toward incorrect answers) contexts. We also find that LLM persuaders significantly increased quiz takers' accuracy, leading to higher earnings, when steering quiz takers toward correct answers, and significantly decreased their accuracy, leading to lower earnings, when steering them toward incorrect answers. Overall, our findings suggest that AI's persuasion capabilities already exceed those of humans that have real-money bonuses tied to performance. Our findings of increasingly capable AI persuaders thus underscore the urgency of emerging alignment and governance frameworks.