Variations On Tree Reconstruction

In a new essay on LessWrong, researcher Adam Shimi presents 'Variations On Tree Reconstruction,' examining how the Genealogical Regularity—the principle that entities descend from predecessors with modifications—enables the Comparative Method across disparate fields. This methodology allows experts to reconstruct ancestral forms by comparing similarities and differences between existing entities. Shimi highlights three major success stories: computational phylogenetics for mapping species evolution, historical linguistics for reconstructing proto-languages, and textual criticism for determining original manuscript versions from error-ridden copies.

While these fields share the core regularity, their implementations differ significantly due to domain-specific factors. Phylogenetics thrives with powerful statistical models, but textual criticism and historical linguistics have not achieved comparable computational success. The key distinction lies in the prevalence of alternative change mechanisms: biology contends with convergent evolution (like carcinization, where unrelated species evolve crab-like features), while linguistics must account for borrowing—words horizontally transferred between languages. These differences determine which analytical methods are portable and effective, revealing both the power and limitations of applying AI and statistical techniques across domains that share structural patterns.