Huang Yiping, Huang Shixin, Chen Xiangjian
Jiangsu University of Science and Technology, Zhenjiang, China.
Yangzhou University, College of Information Engineering, Yangzhou, China.
Sci Rep. 2025 Jul 1;15(1):22133. doi: 10.1038/s41598-025-06280-7.
Enterprises in the context of smart manufacturing face great challenges in terms of human capital strategies as well as incentive mechanisms. Employee Stock Ownership Plans (ESOPs) is one of the key incentive mechanisms with long-term oriented function, but due to the lack of relevant explanations in the context of smart manufacturing, the mechanism of the dynamic impact of ESOPs on corporate performance has not yet been elucidated. In this study, with the idea of combining AI and accounting, we constructed a prediction model of the impact of ESOPs on enterprise performance that integrates language modeling and social sentiment mass data analysis, and introduced the prediction model to analyze the long-term, dynamic and nonlinear impact of ESOPs on enterprises; finally, we constructed an explainable AI (XAI) based on the LSTM model, and used the SHAP value method to explain the impact of ESOPs on enterprise performance. Finally, the Explainable AI (XAI) is built based on the LSTM model, and the SHAP value method is used to downsize the performance of the complex black box model LSTM, present the model "black box", and analyze the common roles played by the elements of ESOPs, the maturity level of smart manufacturing, and the social sentiment on ESOPs in the long term and nonlinear process. Aiming at the above research problems and shortcomings, the main contributions of this paper include: analyzing the dynamic evolution path of ESOP effectiveness from the perspective of intelligent transformation of manufacturing enterprises; predicting the ESOP effectiveness of enterprises through multi-source heterogeneous data (financial data, social sentiment data, operation data) and advanced AI models (LSTM, LLM), and proposing new prediction tools and prediction theories; using XAI technology to realize ESOP effectiveness; and using XAI technology to realize ESOP effectiveness in the long term and non-linear process. Theory; the use of XAI technology to achieve ESOP incentive effect attribution analysis, for management accounting decision support to provide a new dimension of interpretation, which can be used as a research on ESOP dynamic incentive evaluation, integration of non-financial information, predictive analysis of new perspectives for the field of accounting to develop a new research direction, and for the transformation of intelligent manufacturing design and optimization of ESOP to provide empirical data basis and decision support. The study also provides empirical data basis and decision support for the design and optimization of ESOPs during the transformation of smart manufacturing.
在智能制造背景下，企业在人力资本战略和激励机制方面面临巨大挑战。员工持股计划（ESOPs）是具有长期导向功能的关键激励机制之一，但由于在智能制造背景下缺乏相关解释，ESOPs对企业绩效的动态影响机制尚未得到阐明。在本研究中，我们结合人工智能和会计的理念，构建了一个整合语言建模和社会情绪海量数据分析的ESOPs对企业绩效影响的预测模型，并引入该预测模型来分析ESOPs对企业的长期、动态和非线性影响；最后，我们基于长短期记忆（LSTM）模型构建了可解释人工智能（XAI），并使用SHAP值方法来解释ESOPs对企业绩效的影响。最后，基于LSTM模型构建可解释人工智能（XAI），并使用SHAP值方法对复杂黑箱模型LSTM的性能进行降维，呈现模型“黑箱”，分析ESOPs要素、智能制造成熟度水平以及社会对ESOPs的情绪在长期和非线性过程中所起的共同作用。针对上述研究问题和不足，本文的主要贡献包括：从制造企业智能转型的角度分析ESOP有效性的动态演变路径；通过多源异构数据（财务数据、社会情绪数据、运营数据）和先进的人工智能模型（LSTM、语言模型（LLM））预测企业的ESOP有效性，并提出新的预测工具和预测理论；利用XAI技术实现ESOP有效性；以及利用XAI技术在长期和非线性过程中实现ESOP有效性理论；利用XAI技术实现ESOP激励效果归因分析，为管理会计决策支持提供新的解释维度，可作为ESOP动态激励评价、非财务信息整合、预测分析等领域的新视角，为会计领域开拓新的研究方向，并为智能制造转型中ESOP的设计与优化提供实证数据依据和决策支持。该研究还为智能制造转型过程中ESOP的设计与优化提供了实证数据依据和决策支持。
