Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative AI as a Non-Convex Supply Shock: Market Bifurcation and Welfare Analysis

Yukun Zhang, Tianyang Zhang

TL;DR

Generative AI introduces a non-convex supply shock with near-zero marginal cost and a congestion externality from information pollution, challenging convex-SE frameworks. The authors develop a three-layer general equilibrium framework—static vertical differentiation with a kinked frontier, a mean-field dynamic of skill evolution, and a calibrated agent-based simulation—to study transitions. They uncover a middle-class hollow in the skill distribution, a non-monotonic 'Shock Therapy' transition, and an asymmetric skill reconfiguration that favors semantic creativity, with welfare outcomes highly sensitive to pollution intensity. Policy implications emphasize output-side congestion management, including a dynamic Pigouvian tax on AI volume and signal-based human verification to sustain efficient matching in a polluted information environment, highlighting the practical significance for governance in the GenAI era.

Abstract

The diffusion of Generative AI (GenAI) constitutes a supply shock of a fundamentally different nature: while marginal production costs approach zero, content generation creates congestion externalities through information pollution. We develop a three-layer general equilibrium framework to study how this non-convex technology reshapes market structure, transition dynamics, and social welfare. In a static vertical differentiation model, we show that the GenAI cost shock induces a kinked production frontier that bifurcates the market into exit, AI, and human segments, generating a ``middle-class hollow'' in the quality distribution. To analyze adjustment paths, we embed this structure in a mean-field evolutionary system and a calibrated agent-based model with bounded rationality. The transition to the AI-integrated equilibrium is non-monotonic: rather than smooth diffusion, the economy experiences a temporary ecological collapse driven by search frictions and delayed skill adaptation, followed by selective recovery. Survival depends on asymmetric skill reconfiguration, whereby humans retreat from technical execution toward semantic creativity. Finally, we show that the welfare impact of AI adoption is highly sensitive to pollution intensity: low congestion yields monotonic welfare gains, whereas high pollution produces an inverted-U relationship in which further AI expansion reduces total welfare. These results imply that laissez-faire adoption can be inefficient and that optimal governance must shift from input regulation toward output-side congestion management.

Generative AI as a Non-Convex Supply Shock: Market Bifurcation and Welfare Analysis

TL;DR

Generative AI introduces a non-convex supply shock with near-zero marginal cost and a congestion externality from information pollution, challenging convex-SE frameworks. The authors develop a three-layer general equilibrium framework—static vertical differentiation with a kinked frontier, a mean-field dynamic of skill evolution, and a calibrated agent-based simulation—to study transitions. They uncover a middle-class hollow in the skill distribution, a non-monotonic 'Shock Therapy' transition, and an asymmetric skill reconfiguration that favors semantic creativity, with welfare outcomes highly sensitive to pollution intensity. Policy implications emphasize output-side congestion management, including a dynamic Pigouvian tax on AI volume and signal-based human verification to sustain efficient matching in a polluted information environment, highlighting the practical significance for governance in the GenAI era.

Abstract

The diffusion of Generative AI (GenAI) constitutes a supply shock of a fundamentally different nature: while marginal production costs approach zero, content generation creates congestion externalities through information pollution. We develop a three-layer general equilibrium framework to study how this non-convex technology reshapes market structure, transition dynamics, and social welfare. In a static vertical differentiation model, we show that the GenAI cost shock induces a kinked production frontier that bifurcates the market into exit, AI, and human segments, generating a ``middle-class hollow'' in the quality distribution. To analyze adjustment paths, we embed this structure in a mean-field evolutionary system and a calibrated agent-based model with bounded rationality. The transition to the AI-integrated equilibrium is non-monotonic: rather than smooth diffusion, the economy experiences a temporary ecological collapse driven by search frictions and delayed skill adaptation, followed by selective recovery. Survival depends on asymmetric skill reconfiguration, whereby humans retreat from technical execution toward semantic creativity. Finally, we show that the welfare impact of AI adoption is highly sensitive to pollution intensity: low congestion yields monotonic welfare gains, whereas high pollution produces an inverted-U relationship in which further AI expansion reduces total welfare. These results imply that laissez-faire adoption can be inefficient and that optimal governance must shift from input regulation toward output-side congestion management.
Paper Structure (42 sections, 4 theorems, 29 equations, 6 figures, 5 tables)

This paper contains 42 sections, 4 theorems, 29 equations, 6 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Generative AI in Two-Sided Markets
  4. Information Overload and Distribution Dynamics
  5. Welfare and Social Implications
  6. Theoretical Framework: The Analytical Anchor
  7. Demand Side: Endogenous Information Pollution
  8. Supply Side: The Non-Convex Production Frontier
  9. Dynamic Foundations: The Potential Game
  10. Experimental Setup and Calibration
  11. The "Goldilocks" Environment: Micro-Foundations
  12. Bounded Rationality and Adaptive Learning.
  13. Calibration Strategy
  14. Cost Asymmetry ($c_A$ vs $c_H$).
  15. Asymmetric Learning Rates ($\lambda_{tech} \gg \lambda_{creative}$).
  16. ...and 27 more sections

Key Result

Proposition 3.1

There exists a critical noise sensitivity threshold $\beta^*$. For all $\beta > \beta^*$, a technological shock that reduces the marginal cost of AI production ($c_A \to 0$) reduces total social welfare $W$, despite strictly lowering equilibrium prices.

Figures (6)

  • Figure 1: Mechanism: From GenAI cost shock to market bifurcation and welfare trade-offs. A GenAI marginal-cost shock ($c_A \ll c_H$) induces a non-convex production frontier (kink at the AI quality cap $\bar{q}_A$), generating three-tier segmentation (Exit / AI / Human) and a "middle-class hollow" gap in the quality distribution. The expansion of AI volume increases information pollution $\Phi(D_A)$, creating a welfare trade-off between lower prices and higher search costs. These forces feed back into the transition dynamics, producing "shock therapy" (collapse $\rightarrow$ recovery) and asymmetric skill reconfiguration (Tech $\rightarrow$ Creative).
  • Figure 2: Pollution sensitivity reshapes welfare structure. Each panel plots welfare components as a function of AI penetration $D_A$ under different pollution sensitivity levels $\beta$. When pollution sensitivity is low ($\beta = 0.5$), total welfare increases monotonically with AI adoption. As $\beta$ rises, welfare becomes non-monotonic, exhibiting an inverted-U shape. At sufficiently high pollution sensitivity ($\beta \ge 2.5$), the marginal pollution cost dominates price and productivity gains, causing total welfare to decline sharply with further AI expansion.
  • Figure 3: Direct evidence of distribution bifurcation (validation of Lemma \ref{['lemma:hollow']}). Pre-AI, the quality distribution is unimodal and continuous. Post-AI, the supply distribution becomes bimodal with a "hollow zone" (missing middle) around the AI quality cap region, consistent with the non-convex production frontier and the predicted disappearance of mid-tier original content.
  • Figure 4: Static Equilibrium and Welfare Analysis. Top-Left: The S-curve indicates a clean separation of the market, confirming the "Middle-Class Hollow" where mid-tier demand vanishes. Top-Right: Price compression effects. Bottom-Left: Welfare redistribution showing the massive transfer from Producer Surplus to Consumer Surplus.
  • Figure 5: The "Shock Therapy" Transition. Top Panel: Unlike smooth diffusion models, the simulation reveals a near-total collapse of the human supply side (Phase II) before a new equilibrium is established. This validates the hypothesis that information pollution initially overwhelms the market's matching mechanism, leading to a temporary ecological breakdown.
  • ...and 1 more figures

Theorems & Definitions (10)

  • Proposition 3.1: Pollution-Induced Unraveling
  • proof
  • Lemma 3.2: The Middle-Class Hollow
  • proof
  • Proposition 3.3: Market Segmentation
  • proof
  • Theorem 3.4: Local Stability via Free Energy
  • proof
  • proof
  • proof