This study focused on three main entities involved in industry-education integration: Universities, enterprises, and the government. Based on the evolutionary game theory, a tripartite evolutionary game model was constructed to thoroughly evaluate the dynamics between interest evolutionary game relationships and strategic choices among these entities during the process of industry-education integration. By setting six strategies such as “active cooperation” and “passive cooperation” for universities, “deep participation” and “formal participation” for enterprises, and “regular supervision” and “passive supervision” for the government, the study systematically analyzed the stability conditions for each entity’s strategic choices through a payoff matrix and replication dynamic equations. It was found that the stability of strategic choices for universities, enterprises, and the government depended on the comparison of net benefits under different strategies. When the net benefits of active cooperation, deep participation, and regular supervision exceeded those of their respective passive strategies, each entity tended to choose the active strategy owing to its stability. Otherwise, they would lean towards passive strategies. This study revealed the inherent laws governing the strategic choices of the three entities in industry-education integration, thus providing a theoretical basis and policy recommendations for optimizing integration policies and promoting tripartite cooperation. It holds significant importance for driving high-quality development in the proposed industry-education integration.
The proliferation of Large Language Models (LLMs) presents systemic challenges, including the generation of misinformation, hallucinations, and amplification of societal biases, causing a broader “epistemic crisis”. This paper introduced the novel Synergistic Algorithmic Repair Framework designed to mitigate the algorithmic harm. The methodology integrated three components into a continuous feedback loop: (1) The proactive curation of a verifiable digital ecosystem to establish ground truth; (2) the systematic application of Verifiable Human Feedback to correct LLM output with evidence-based inputs; and (3) the strategic curation of verified data into datasets for long-term model refinement. The efficacy of the framework was assessed through a qualitative case study involving two distinct real-world applications. The results demonstrated that the intervention successfully suppressed the targeted disinformation campaign of search engines and transformed LLM-generated output from negative or non-existent to positive or factual, in order to align with the curated ground truth. The study concluded that the synergistic operation of these components provided a durable and resilient method for repairing algorithmic harm. This framework, as a practical tool, offered promising insights for organizations and individuals to achieve digital equity and served as an operational model for implementing principles of ethical AI governance, such as human oversight and accountability.
Open Access
Journal Metrics
Journal Overview
Associated Journals