Author Network Centrality Drives Citation Disparities in AI Conferences
Analysis
This paper investigates how the position of authors within collaboration networks influences citation counts in top AI conferences. It moves beyond content-based evaluation by analyzing author centrality metrics and their impact on citation disparities. The study's methodological advancements, including the use of beta regression and a novel centrality metric (HCTCD), are significant. The findings highlight the importance of long-term centrality and team-level network connectivity in predicting citation success, challenging traditional evaluation methods and advocating for network-aware assessment frameworks.
Key Takeaways
- •Author network centrality significantly impacts citation counts in top AI conferences.
- •Long-term centrality metrics are more predictive of citation success than short-term ones.
- •Team-level network connectivity is crucial for explaining citation variance.
- •The study proposes a novel centrality metric (HCTCD) and uses beta regression for citation analysis.
- •Integrating centrality features improves citation prediction accuracy, suggesting the need for network-aware evaluation frameworks.
“Long-term centrality exerts a significantly stronger effect on citation percentiles than short-term metrics, with closeness centrality and HCTCD emerging as the most potent predictors.”