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

Carlos J. Costa,Joao Tiago Aparicio,Manuela Aparicio

2024-11-14

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.

Computers and Society

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: **How to predict and evaluate the impact of generative artificial intelligence (Generative AI) on the economy and society**. With the wide application of generative AI, it not only changes the technological landscape but also profoundly reshapes the social structure. Although generative AI brings huge innovation potential, it also comes with a great deal of uncertainty and complexity. Therefore, there is an urgent need to develop robust and scalable analysis frameworks to comprehensively understand its impact.

Specifically, through a comprehensive literature review, the paper identifies and evaluates a variety of methodologies aimed at helping to predict the economic and social consequences of generative AI adoption. These methods include:

1. **Agent - Based Simulation (ABS)**:
- **Advantages**: It can simulate individual entities and their interactions and explore complex dynamic behavior patterns.
- **Disadvantages**: It is data - intensive, depends on model assumptions, and is relatively complex.

2. **Econometric Models**:
- **Advantages**: It utilizes historical data and provides a robust framework to analyze the relationships between economic variables.
- **Disadvantages**: It has simplified assumptions, may be inaccurate, and is mainly limited to the economic aspect.

3. **Input - Output Analysis**:
- **Advantages**: It quantifies the interdependencies between different economic sectors and predicts production changes.
- **Disadvantages**: It is limited to economic impacts, ignores social / cultural aspects, and has high data requirements.

4. **Reinforcement Learning (RL)**:
- **Advantages**: It adapts to changing environments and predicts the evolution of system - human interactions.
- **Disadvantages**: It requires a large amount of training data, has limited generalization ability, and may be biased.

5. **Surveys and Interviews**:
- **Advantages**: It directly collects opinions and understands human perception.
- **Disadvantages**: It has limited representativeness, is prone to bias, and is not suitable for long - term prediction.

6. **Scenario Analysis**:
- **Advantages**: It explores different future scenarios and is helpful for strategic planning.
- **Disadvantages**: It depends on the quality of scenarios and cannot provide accurate predictions.

7. **Policy Analysis**:
- **Advantages**: It evaluates policy effects and supports decision - making.
- **Disadvantages**: It focuses on specific policies, may overlook broader changes, and requires comprehensive policy options.

8. **Delphi Method**:
- **Advantages**: It is a structured collection of expert opinions to reach a consensus.
- **Disadvantages**: It is highly subjective, is easily influenced by expert biases, and is not suitable for specific predictions.

Through the evaluation of these methods, the paper aims to provide valuable insights for decision - makers and policy - makers, helping them make wise decisions in an era of unprecedented technological change, ensuring responsible innovation and promoting economic growth.

In summary, through systematically evaluating various methodologies, this paper addresses the complexity and uncertainty issues brought by generative AI in the economic and social fields, providing multi - dimensional analysis tools to meet the challenges brought by this technological revolution.