Kun Tan

Knowledge-intensive organizations in the energy sector have long grappled with the “knowledge-sharing prisoner's dilemma,” where individual motivations to withhold achievements lead to collective inefficiency and data silos. When individuals conceal knowledge driven by short-term incentives, the entire organization bears systemic costs—including redundant tasks, stalled innovation processes, exacerbated data silos, and structural challenges in talent development and retention.

To fundamentally reverse this negative cycle, this report innovatively proposes building a “positive feedback flywheel for research sharing” powered by large language models (LLMs). This flywheel is driven by three core algorithms: contribution rewards based on Shapley values, LLM credibility verification, and GNN-driven intelligent recommendations. This approach does not merely treat LLMs as productivity tools but deeply embeds them as intelligent infrastructure that propels organizational culture from ‘hoarding’ to “sharing.” This flywheel mechanism operates through three interconnected phases: First, LLMs function as intelligent R&D assistants, significantly accelerating individual researcher efficiency and delivering immediate, tangible rewards. Second, this enhanced efficiency motivates researchers to share their data and knowledge via the platform, thereby earning quantifiable recognition and rewards for their contributions. Finally, the vast amounts of high-value data generated through sharing continuously feed back into the large model, creating powerful network effects that attract more participants and contributions, ultimately achieving exponential growth in research efficiency.

The efficacy of this intervention will be empirically validated through a Stepped-Wedge Cluster Randomized Trial (SW-CRT) involving 14 research departments in China Petroleum Safety and Environmental Protection Technology Research Institute. By altering the game's payoff matrix from competition to cooperation, this algorithm-driven model offers a replicable blueprint for energy organizations to unlock collective intelligence and accelerate innovation in energy safety and energy research.

Co-author/s:

Kai Zhang, Professor, China University of Petroleum.