An Economic Perspective of Explainable Artificial Intelligence on Intellectual Property
Abstract
This research explores the economic aspects of Explainable Artificial Intelligence (XAI) in Intellectual Property (IP) management, leveraging sophisticated machine learning techniques to measure the potential of the economic gain from XAI including, improved valuation accuracy for IP, increased revenue from licensing, and decreased costs when litigating. We used a global patent dataset including firm-level financial metrics and governance data. Predictive models including Gradient Boosted Decision Trees (GBDT), Explainable Boosting Machines (EBM), and Random Forests (RF) were developed. The economic benefits were evaluated through a SHAP-based feature importance analysis which identified the key factors for economic gain was the Explainability Score, Digital Transparency Index, and Market Capitalization. Overall, GBDT achieved the highest accuracy ($\mathrm{R}^{2}=0.95$; MAPE = 4.8%) followed by EBM ($\mathrm{R}^{2}=0.92$) which provided more meaningful interpretability. Compared to existing studies, employing XAI meant that IP portfolios were now better able to maximize economic performance compared to traditional AI models. Our study contributes to the existing literature by combining economic theory with legal considerations and explainable machine learning, and by providing tangible guidance for policymakers, corporate strategists, and regulators of technological innovations.