AI-Driven Research Analytics and Policy: Applications of Network Analysis, GNNs, and LLMs

Research metrics and analytics are pivotal for evaluating scientific impact, informing policy, and guiding strategic management in an era of exponential knowledge growth. Traditional methods like citation counts and h-indices face limitations in capturing complex, relational dynamics of scholarly communication, such as collaboration networks, knowledge flows, and semantic relationships. Recent advances in artificial intelligence, particularly network analysis for mapping influence, graph neural networks (GNNs) for heterogeneous scholarly graphs, and large language models (LLMs) for processing vast textual corpora, offer transformative potential. United by their capacity to operate on the complementary structured and unstructured data of scholarly ecosystems (citation graphs, co-authorship networks, and full-text corpora) these methods together enable a multi-modal understanding of research landscapes that no single approach can achieve alone. These technologies enable nuanced insights into research ecosystems, from predicting breakthroughs to assessing interdisciplinary impact and informing R&D funding decisions. As research output surges, driven by global collaboration and open science, integrating these AI methods into metrics is essential for robust policy and strategic decisions.This Research Topic addresses the gap between static metrics and the dynamic, multifaceted nature of modern science by advancing technical development and implementation of AI-driven research analytics. Current challenges include scalability in handling massive citation and co-authorship graphs, interpretability of black-box models in policy contexts, and ethical integration of LLMs for bias-free evaluation. To achieve breakthroughs, we seek contributions leveraging recent advances: GNN architectures (e.g., GraphSAGE, GAT, and heterogeneous GNN variants) for node embeddings in Altmetrics; network motifs for detecting innovation hotspots; LLM fine-tuning (e.g., via PEFT) for automated summarization and trend forecasting; and further integration of those methods to handle complicated downstream tasks. Outcomes include validated frameworks for real-world deployment, benchmarks comparing AI vs. classical methods, and policy guidelines for R&D evaluation and strategic research management. This will empower stakeholders to navigate 2026’s AI-accelerated research landscape with precision.This Research Topic explores the technical development and implementation of advanced AI methods, such as network analysis, graph neural networks (GNNs), and large language models (LLMs), to revolutionize research metrics and analytics in scientometrics and bibliometrics. It aims to bridge cutting-edge computational techniques with practical applications in scholarly evaluation, policy formulation, and strategic innovation management, addressing the limitations of traditional metrics amid surging global research output. Suggested themes which may be explored in this Topic include, but are not limited to:• GNN applications in scholarly knowledge graphs, citation networks, and impact prediction.• LLM-enhanced topic modelling, semantic analysis of abstracts/preprints, and trend forecasting.• Hybrid network-LLM models for research evaluation, collaboration analysis, R&D portfolio assessment, and policy simulations.• Scalable implementations of AI analytics for big scholarly data, including dynamic graphs.• Ethical considerations, bias mitigation, and validation studies in AI-driven bibliometricsWe welcome Original Research (empirical/technical papers), Methods (novel algorithms/tools), Policy and Practice Reviews (implementation case studies), and Perspectives (future visions). Manuscripts should demonstrate rigorous evaluation and discuss policy/strategic implications. Topic editors: Mengjia Wu Source: https://www.frontiersin.org/research-topics/79827/ai-driven-research-analytics-and-policy-applications-of-network-analysis-gnns-and-llms