Mediating Similarity: An Information-Theoretic Principle of Reference Behavior

Area of Science:

• Bibliometrics
• Information Science
• Network Science

Background:

• Information theory quantifies knowledge combinations, but reference selection principles in science are understudied.
• Journal citation networks offer a large-scale data source to investigate scientific communication strategies.

Purpose of the Study:

• To introduce and empirically validate the "Mediating Similarity" principle in scientific referencing.
• To analyze how journals use reference lists as strategic cognitive bridges for information dissemination.
• To explore the mechanisms and universality of this referencing principle.

Main Methods:

• Analysis of a large-scale journal citation network.
• Application of information-theoretic measures and computational experiments.
• Perturbation experiments to test reference selection mechanisms.

Main Results:

• The Mediating Similarity principle is universally applicable, showing mediated paths via references are more efficient than direct paths.
• Reference selection involves a dual mechanism: synergistic optimization and "satisficing" to balance ideal paths with practical constraints.
• Journals' reference portfolios are not random but strategically curated combinations.

Conclusions:

• Reference behavior in science is a strategic process of navigating cognitive landscapes.
• Journals selectively curate references to enhance integrative capacity and innovative potential.
• The study reframes scientific referencing as a key factor in knowledge discovery and integration.