The Spillover Effects of Peer AI Rinsing on Corporate Green Innovation

The Spillover Effects of Peer AI Rinsing on Corporate Green Innovation

Overview

This study investigates the impact of "AI washing" - the practice of using artificial intelligence merely as cosmetic embellishment rather than a genuine driver of transformation - on corporate green innovation in China. Using natural language processing techniques to analyze annual reports, the researchers find that:

• Corporate AI washing exerts a significant crowding-out effect on green innovation.
• This negative relationship is transmitted through dual mechanisms in product and capital markets.
• The crowding-out effect exhibits heterogeneity, with private enterprises, SMEs, and firms in highly competitive industries most adversely impacted.
• Simulation results indicate that a combination of policy tools - enhanced disclosure regulation, consumer education, and reputational sanctions - can effectively improve market equilibrium.

Problem & Context

• Artificial intelligence (AI) is seen as a core driver of the Fourth Industrial Revolution in China, while green technological innovation is key to achieving the country's "dual carbon" goals.
• However, "AI washing" - the use of AI as cosmetic embellishment rather than genuine transformation - has become a systemic risk, creating a disconnect between disclosed content and actual applications.
• This AI washing behavior exhibits contagious spread within industries, exacerbating market information asymmetry and potentially crowding out green innovation.

Methodology

1. Measurement of AI Washing:
   • Used large language models to classify AI-related statements in annual reports as "descriptive" (lacking technical details) or "substantive" (objective inputs/outputs).
   • Constructed firm-level AI washing index as the proportion of descriptive statements.
   • Also calculated industry-level peer AI washing index.
2. Empirical Analysis:
   • Used two-way fixed effects, IV, and dynamic panel GMM models to examine the impact of AI washing on green innovation.
   • Tested product market (market share) and capital market (financing constraints) mechanisms.
   • Conducted heterogeneity analyses by ownership, firm size, and industry competition.
3. Agent-based Simulation:
   • Developed an artificial market model to simulate the dynamic evolution of AI washing and green innovation.
   • Tested baseline scenario and policy intervention scenarios (enhanced regulation, consumer education, reputational sanctions).

Results

1. Empirical Findings:
   • Peer AI washing exerts a significant negative effect on firms' green innovation output.
   • This effect is transmitted through both product market (market share erosion) and capital market (financing constraints) channels.
   • Private enterprises, SMEs, and firms in highly competitive industries are most adversely impacted.
2. Simulation Findings:
   • Baseline scenario shows a "bad money drives out good" dynamic, with AI washing rapidly spreading and green innovation declining.
   • Policy interventions, especially a combination of enhanced disclosure, consumer education, and reputational sanctions, can effectively curb AI washing and improve green innovation.

Interpretation

• AI washing creates a "lemons market" where genuine green innovators are crowded out, undermining overall innovation in the industry.
• Heterogeneous effects suggest private firms, SMEs, and competitive industries are most vulnerable to the negative spillovers of AI washing.
• Comprehensive policy solutions targeting both supply-side (disclosure) and demand-side (market recognition) are needed to reshape the market equilibrium.

Limitations & Uncertainties

• Reliance on textual analysis of annual reports to measure AI washing, despite efforts to validate the methodology.
• Potential endogeneity issues, though addressed through IV and dynamic panel approaches.
• Simplifying assumptions in the agent-based simulation, though sensitivity analyses support the robustness of the conclusions.

What Comes Next

• Further research on the heterogeneous impacts of AI washing across industries, regions, and firm lifecycles.
• Exploration of alternative policy instruments, such as targeted tax incentives or public procurement preferences, to incentivize genuine AI-enabled green innovation.
• Investigation of the potential long-term strategic responses of firms to combat the negative effects of AI washing, such as greater vertical integration or diversification.