国内统一连续出版物号:CN 11-1384/F

国际标准连续出版物号:ISSN 1000-7636

当前位置: 首页  >>   最新刊发  >>   最新刊发
最新刊发

人工智能重塑创新绩效:机制解构与实践探索

人工智能重塑创新绩效:机制解构与实践探索

胡朝阳1 张自然2

(1.中国社会科学院大学;2.中国社会科学院)

摘要:人工智能(AI)正深度融入中国创新体系,其微观机制亟须结合本土实践加以探究。本文拓展离散时间无限期优化模型,将AI赋能水平内生嵌入知识演化、创新成本、创新产出与市场收益函数,构建赋能、降本与激励效应的统一分析框架,验证三重机制的存在性及交互影响。在此基础上,选取科大讯飞公司作为典型案例,剖析机制协同及“AI漂洗”风险防范路径,实现理论与实践的交叉验证。理论推导发现:过度激励易诱发“AI漂洗”,进而弱化赋能效应与降本效应。其中,赋能效应着眼于创新弹性与转化效率的提升;降本效应关注压缩结构性成本的筛选;激励效应聚焦信号传递及其引发的融资溢价。本文的案例研究揭示了三重机制在企业实践中的动态耦合,精准契合创新主体通过AI赋能、AI降本与AI推动的融资激励驱动创新绩效提升的核心诉求,并识别出战略自律、财务透明与生态约束等防范“AI漂洗”风险的有效机制。本文为理解AI如何提升创新绩效提供统一理论解释与本土实证证据,亦为完善AI相关政策设计提供了有益启示。

关键词:人工智能;创新绩效;赋能效应;降本效应;激励效应;AI漂洗

作者简介:胡朝阳,中国社会科学院大学经济学院博士研究生,北京,102488;张自然,中国社会科学院经济研究所研究员,博士生导师,通信作者,北京,100836。

基金项目:国家社会科学基金重点项目“降碳、减污、扩绿、增长协同推进的理论与实践研究”(24AZD074)

引用格式:胡朝阳,张自然. 人工智能重塑创新绩效:机制解构与实践探索[J]. 经济与管理研究,2026,47(11):61-75.


Artificial Intelligence Reshaping Innovation Performance: Mechanism Deconstruction and Practical Exploration

HU Zhaoyang1, ZHANG Ziran2

(1. University of Chinese Academy of Social Sciences, Beijing 102488;

2. Chinese Academy of Social Sciences, Beijing 100836)

Abstract: Against the backdrop of artificial intelligence (AI) being deeply integrated into national innovation systems to drive high-quality economic development, clarifying the micro-level mechanisms and local practical logic through which AI reshapes innovation performance holds significant theoretical and practical importance. This paper aims to open the “black box” of AI-driven innovation and systematically investigate its internal mechanisms, interactive effects, and risk prevention pathways. Methodologically, this paper extends a discrete-time infinite-horizon dynamic optimization model by endogenously embedding the AI empowerment level into knowledge evolution equations, nonlinear cost functions, and market revenue signal generation mechanisms. A unified intertemporal dynamic analytical framework encompassing empowerment, cost reduction, and incentives is constructed, and the existence of these triple mechanisms, along with their interactive impacts, is rigorously verified via mathematical deduction. Building on this theoretical foundation, iFlytek Co., Ltd. is selected as a typical case to analyze the dynamic coupling of these mechanisms in corporate practice and the prevention pathways for “AI washing” risks, thereby achieving cross-validation between theoretical deductions and empirical case findings.

The results are threefold. First, AI enhances knowledge conversion efficiency per unit of effort and innovation output elasticity through the empowerment effect, boosting the full-cycle marginal returns of innovation activities. Second, AI systematically compresses trial-and-error costs and risk exposure through the cost reduction effect, restructuring the marginal cost structure of innovation. Third, AI transforms technology adoption behaviors into credible signals of high technical adaptive capacity through the incentive effect, triggering capital market financing premiums and forming a positive feedback loop of signal recognition and resource reinvestment. More critically, this paper reveals complex interactive relationships among these mechanisms: excessive incentives can easily trigger “AI washing”, which in turn weakens the empowerment and cost reduction effects through resource crowding out and allocation distortions, thereby undermining long-term innovation capacity. The case analysis not only validates the synergistic role of the triple mechanisms but also identifies a risk prevention mechanism centered on strategic self-discipline, financial transparency, and ecological constraints.

This paper achieves logical integration and mathematical consistency of the triple mechanisms within a unified framework, revealing the dual attributes of AI as both a factor of production and a vehicle of capability signals. The conclusions provide systematic theoretical explanations and local empirical support for understanding how AI enhances innovation performance, while offering crucial decision-making insights for innovators to avoid false publicity traps, enterprises to build compliance and risk control systems, and policymakers to improve capital market signal screening mechanisms and advance the implementation of the AI Plus initiative.

Keywords: artificial intelligence; innovation performance; empowerment effect; cost reduction effect; incentive effect; AI washing


下载全文