C3-index: A PageRank based multi-faceted metric for authors' performance measurement

Dinesh Pradhan, Partha Sarathi Paul, Umesh Maheswari, Subrata Nandi, Tanmoy Chakraborty

The academic landscape is often governed by the "publish or perish" mantra, where metrics like the h-index dictate career trajectories. However, traditional metrics frequently fail to distinguish between researchers at similar career stages or identify emerging talent. This paper introduces the $C^3$-index, a PageRank-based metric that utilizes multi-layered networks to provide a more granular and predictive assessment of scientific impact.

The Resolution Problem in Academic Metrics

A primary criticism of the h-index is its "resolution" problem. Because it relies on integer values, thousands of researchers often share the same score, creating massive ties. This makes it difficult for hiring committees to differentiate between the majority of scientists who are not yet "superstars."

Distributions of h-index and g-index (a, c) and their percentage counts by year (b, d), showing the heavy concentration of authors with low scores.

As shown in the distributions above, a significant percentage of authors cluster at the lower end of the h-index and g-index scales. For instance, in 1998, nearly 80% of authors had an h-index of zero. Furthermore, these metrics are "lagging indicators"—they reward established scientists for past achievements rather than identifying promising young researchers at the start of their careers.

Methodology: Beyond Simple Citation Counts

The $C^3$-index moves from measuring simple popularity (citation counts) to measuring prestige (the influence of those citing and collaborating with an author). It achieves this by constructing a weighted multi-layered network consisting of three layers:

  1. Author Citation Layer: A directed graph of authors citing other authors.
  2. Author Coauthorship Layer: An undirected graph of collaboration.
  3. Paper Citation Layer: The standard network of papers citing papers.

By applying a modified PageRank algorithm across these layers, the $C^3$-index calculates a composite score based on three sub-indices: the Author Citation Index (ACI), the Paper Citation Index (PCI), and the Author coAuthorship Index (AAI). This approach recognizes that a citation from a Nobel laureate carries more "prestige" than one from a low-impact source, and that high-performing researchers often collaborate with one another.

Key Insights and Results

The researchers tested the $C^3$-index using a massive dataset from Microsoft Academic Search, encompassing 6.6 million papers and 3.1 million authors in Computer Science. Key findings include:

  • Breaking Ties: Unlike the h-index, the $C^3$-index provides a continuous decimal score, effectively resolving the "tie" issue for researchers with similar citation counts.
  • Predictive Power: The $C^3$-index acts as a "leading indicator." The authors found that researchers who eventually reached high h-indexes could have been identified by their high $C^3$-index scores 4 to 5 years earlier.
  • Low Correlation with h-index: The metric shows only a 60% correlation with the h-index, suggesting it captures unique information—specifically regarding collaboration quality—that traditional metrics ignore.

Conclusion

The $C^3$-index represents a significant step forward in bibliometrics by shifting the focus from quantity to the quality of interactions within the scientific community. While currently focused on Computer Science, its ability to identify "scientific stars" years in advance makes it a potentially invaluable tool for academic evaluation, tenure decisions, and resource allocation.

Summary updated

BibTeX
@article{pradhan2017c,
  author = {Dinesh Pradhan and Partha Sarathi Paul and Umesh Maheswari and Subrata Nandi and Tanmoy Chakraborty},
  title = {{C3-index: A PageRank based multi-faceted metric for authors' performance measurement}},
  journal = {Scientometrics},
  year = {2017},
  doi = {10.1007/s11192-016-2168-y},
}
Figure