Recommending Under-Represented Influential Researchers Using Geographically Aware Contrastive Learning

Academic recognition and leadership roles in research communities do not always reflect the full spectrum of scholarly influence. Our analysis identifies a subset of influential authors whose citations span many geographic regions; we refer to them as geographically inclusive influencers. Despite having comparable overall citation impact, they are systematically underrepresented in leadership positions relative to peers whose influence is regionally concentrated. To address this disparity, we introduce GeoGCL (Geographic-aware Graph Contrastive Learning), a representation learning framework that jointly models structural influence and geographic inclusivity in citation networks. GeoGCL combines a PNA-based encoder with a novel inclusivity-aware contrastive objective to learn node embeddings that reflect both multi-hop influence pathways and cross-regional citation reach. These embeddings are used in a greedy selection strategy to recommend top-k influencers that maximize influence spread while promoting geographic diversity. Evaluated on AMiner and DBLP datasets, GeoGCL outperforms existing fairness-aware influence maximization baselines in both influence spread and inclusivity, without sacrificing scholarly impact. Our work offers a scalable, embedding-based solution to surface underrepresented yet influential scholars, supporting more equitable academic recognition.