A note and a correction on measuring cognitive distance in multiple dimensions

TL;DR

This paper clarifies and corrects methods for measuring cognitive distance in multiple dimensions, emphasizing the importance of scale invariance and comparing different approaches that yield similar results.

Contribution

It introduces the term 'similarity-adapted publication vectors', corrects previous normalization errors, and compares three methods for measuring cognitive distance.

Findings

All three approaches produce very similar results.

Corrected normalization improves measurement accuracy.

Emphasizes importance of scale invariance in cognitive distance measures.

Abstract

In a previous article (Rahman, Guns, Rousseau, and Engels, 2015) we described several approaches to determine the cognitive distance between two units. One of these approaches was based on what we called barycenters in N dimensions. The present note corrects this terminology and introduces the more adequate term 'similarity-adapted publication vectors'. Furthermore, we correct an error in normalization and explain the importance of scale invariance in determining cognitive distance. We also consider weighted cosine similarity as an alternative approach to determine cognitive (dis)similarity. Overall, we find that the three approaches (distance between barycenters, distance between similarity-adapted publication vectors, and weighted cosine similarity) yield very similar results.