Leveraging Phylogenetics in Software Product Families: The Case of Latent Content Generation in Video Games

Abstract

A family of software products comprises similar products within a defined scope that share common characteristics, often due to reuse techniques applied during development. This paper introduces an approach that applies biological insights to map the landscape of a software product family, identifying potential gaps within its scope. Phylogenetics studies the gene similarity among groups of organisms to understand ancestry among species. Leveraging Phylogenetics in software, our approach offers a structured view of a product family, aiding in the discovery of unexplored areas fitting the scope of the family. Our approach creates a phylogenetic tree that enables to easily identify latent products (ancestors) that did not exist in the original family. Those ancestors can then be reconstructed from existing products (descendants). The product family evaluated is a set of industry-scale video game non-playable characters. We assess this approach through video game simulations and scope metrics to determine how closely the reconstructed products align with the family's scope. The results confirm that the content generated with phylogenetics aligns better with the family scope than the state-of-the-art procedural content generation techniques using evolutionary algorithms. Phylogenetics enhances content generation by providing a framework to understand and expand the product family with new content.