Socio-Economic Consequences of Generative AI: A Review of Methodological Approaches

TL;DR

This paper reviews various methodological approaches used to predict the economic and social impacts of generative AI, highlighting their strengths, weaknesses, and suitability for addressing uncertainty and resource constraints.

Contribution

It provides a comprehensive overview of existing methodologies for assessing generative AI's societal impacts, comparing their effectiveness and limitations.

Findings

Agent-Based Simulation effectively models complex social interactions.

Econometric Models are useful for quantitative impact analysis.

Surveys and Delphi methods gather expert opinions on policy implications.

Abstract

The widespread adoption of generative artificial intelligence (AI) has fundamentally transformed technological landscapes and societal structures in recent years. Our objective is to identify the primary methodologies that may be used to help predict the economic and social impacts of generative AI adoption. Through a comprehensive literature review, we uncover a range of methodologies poised to assess the multifaceted impacts of this technological revolution. We explore Agent-Based Simulation (ABS), Econometric Models, Input-Output Analysis, Reinforcement Learning (RL) for Decision-Making Agents, Surveys and Interviews, Scenario Analysis, Policy Analysis, and the Delphi Method. Our findings have allowed us to identify these approaches' main strengths and weaknesses and their adequacy in coping with uncertainty, robustness, and resource requirements.