Artificial Intelligence as a Meta Good: Dynamic Properties and Economic Implications
Article Date: 2025-11-06T05:16:07+00:00
Summary
Fadi Fawaz develops a theoretical framework that classifies AI as both a “meta good” and a “dynamic good.” A meta good is one that generates downstream outputs (text, code, images) and thus expands what consumers can use; a dynamic good improves endogenously through use, feedback and retraining. The paper builds a welfare-theoretic model showing that when learning-by-using and data externalities exist, decentralised markets underprovide AI adoption relative to the social optimum. Formal propositions set out the taxonomy and welfare ranking; a stylised calibration shows diffusion and quality dynamics consistent with the model. The framework is presented as generalisable to other general-purpose technologies and is tied to contemporary policy debates on governance, openness, and international coordination for equitable diffusion.
Key Points
- Defines two new categories: “meta goods” (produce downstream outputs that expand feasible consumption) and “dynamic goods” (quality improves with use and feedback).
- Argues AI exhibits both properties simultaneously, which changes standard classifications of economic goods.
- Develops a welfare-theoretic model showing decentralised equilibria can underprovide AI adoption when learning-by-using and data externalities are present.
- Shows a social planner would favour higher adoption or coordination to internalise data and learning externalities, improving welfare.
- Provides a stylised calibration illustrating realistic diffusion and quality trajectories consistent with the theory.
- Highlights policy implications: debates on open vs closed access, data sharing, subsidies, and international institutional roles are central to equitable AI diffusion.
- Framework extends beyond AI to other general-purpose technologies with similar meta/dynamic properties.
Context and relevance
This paper speaks directly to current policy and economic discussions about how AI should be governed and spread across societies. By reframing AI as both a meta and a dynamic good, it clarifies why simple market outcomes may leave societies with too little adoption or uneven quality improvements. The results are relevant to regulators, international bodies, economists, and corporate strategists thinking about data governance, public investment, and incentives for sharing or opening AI systems.
Author style
Punchy and analytical: the author combines clear definitions with formal welfare analysis and a simple calibration. If you care about why market incentives alone may not deliver broad, high-quality AI diffusion, this paper makes the case crisply and rigorously.
Why should I read this?
Look — if you want a tidy way to think about why AI feels different from previous technologies and what that means for policy, this paper is worth 15 minutes. It gives you a neat label (meta + dynamic) and shows the practical sting: without coordination or policy fixes, markets may underdeliver the adoption and quality improvements we actually want.
Source
Source: https://onlinelibrary.wiley.com/doi/10.1111/ajes.70011?af=R
