Towards Intelligent Research Collaboration: A Hybrid AI Framework for Recommending Participants in Research Projects

The success of research project proposals largely depends on the quality of the consortium, which must possess strong expertise and experience aligned with the themes of the relevant funding calls, such as those under the EU’s Horizon Europe programme. However, forming such a consortium remains one of the most difficult tasks, as it involves identifying suitable research collaborators. Traditional approaches typically rely on social networks or citation metrics, but these have shown limited effectiveness. This paper introduces an Agentic Graph-based Retrieval-Augmented Generation (RAG) approach that delivers contextualized and explainable collaborator recommendations, tailored to researchers’ expertise and the relevance of proposed projects, offering improved performance over conventional methods. The approach integrates the strengths of Knowledge Graphs (KGs) and Large Language Models (LLMs), and has been developed using the Design Science research methodology. Its effectiveness was assessed using two of the top-performing LLMs currently available: Claude Sonnet 3.5 and GPT-4o.