The AI Augmented Macro Economy: A New-Keynesian DSGE Model with AI Driven Expectations Formation

Abstract

The combination of Artificial Intelligence (AI) and Machine Learning (ML) in business and economics is a game changer, similar to the effect that Industrial Revolution had on production. This paper studies the macroeconomic consequences of such a world in which a large fraction of economic agents – from firms to households – use increasingly sophisticated AI models for forecasting and decision-making. We construct a New-Keynesian DSGE model that explicitly considers rational expectations agents and AI-informed forecasters. The AIs utilize a non-parametric, flexible methodology that resembles some of the modern ML algorithms to generate anticipations on future macro variables such as inflation and output. Our model suggests that AI agents do a better job understanding the structure of the economy – including its non-linearity and the dynamics of policy transmission – than humans. We derive the steady state of the model, and linearize it to characterize its impulse response functions. Our results provide evidence that an economy staffed with AI-generated forecasters features much more pronounced business cycle dynamics, in response to pure technology (productivity) shocks. In the presence of unexpected aggregate demand or policy shocks, however, AI agents lead to speedier macroeconomic stabilization because their forecasts react more rapidly to new information regimes. The implications for policymakers are discussed in this paper, emphasizing the necessity for central banks to question their monetary policy frameworks when faced with new and more intelligent economic agents as well as potential dangers of "herding" behavior among similar AI forecasters.