A diversity metric that accounts for (dis)similarity of entities

What is it about?

Traditionally, diversity has been quantified in terms of a richness component and an abundance component with richness measuring the number of distinct entity types (breadth) and the abundance accounting for the quantity (depth) of each entity type. However, an often ignored aspect of diversity is the amount of dissimilarity between the entities. In this work, we propose a way of measuring diversity that accounts for the dissimilarity among entities. We use a hierarchical distance tree with the entities as leaf nodes of the tree, common attributes among the entities as the parent nodes, and use the average distance between the leaf nodes as the measure of diversity. To develop the tree and identify all the common attributes and the relationship between them, we engage the domain experts in a card sorting exercise. We use this metric (DSCORE) to capture the diversity of activity types in an online user community.

Featured Image

Photo by Victoriano Izquierdo on Unsplash

Photo by Victoriano Izquierdo on Unsplash

Why is it important?

Measuring diversity of species, entities, objects, particles, or variables in a system becomes important in a number of contexts. However, it is important to choose the correct metric that captures the diversity the best. Traditional diversity measures such as Richness, Shannon Entropy, Simpson Index, or the Gini Simpson Index that use richness and abundance are not always useful or desirable. For instance, we all know that a zoo with five different types of primates is not as diverse as a zoo with a chimpanzee, a whale, and a snake. This example demonstrates that taking (dis)similarity among the entities also into consideration captures diversity the best

Perspectives