Computational modelling of evolution: ecosystems and language

Recently, computational modelling became a very important research tool that enables us to study problems that for decades evaded scientific analysis. Evolutionary systems are certainly examples of such problems: they are composed of many units that might reproduce, diffuse, mutate, die, or in some cases for example communicate. These processes might be of some adaptive value, they influence each other and occur on various time scales. That is why such systems are so difficult to study. In this paper we briefly review some computational approaches, as well as our contributions, to the evolution of ecosystems and language. We start from Lotka-Volterra equations and the modelling of simple two-species prey-predator systems. Such systems are canonical example for studying oscillatory behaviour in competitive populations. Then we describe various approaches to study long-term evolution of multi-species ecosystems. We emphasize the need to use models that take into account both ecological and evolutionary processes. Finally, we address the problem of the emergence and development of language. It is becoming more and more evident that any theory of language origin and development must be consistent with darwinian principles of evolution. Consequently, a number of techniques developed for modelling evolution of complex ecosystems are being applied to the problem of language. We briefly review some of these approaches.

💡 Research Summary

The paper provides a broad review of computational approaches to modeling evolutionary dynamics, focusing on two major domains: ecological systems and language evolution. It begins with a discussion of classical Lotka‑Volterra equations for two‑species predator‑prey interactions, pointing out their limitations: sensitivity to initial conditions, structural instability, and the lack of spatial heterogeneity. The authors then argue for individual‑based (agent‑based) models that operate on a lattice, where each site can be empty, occupied by a prey, a predator, or both. A single control parameter r determines the probability of updating a prey or a predator. Simulations reveal a phase transition: for sufficiently large r both species coexist (active phase), while for smaller r predators die out and the system falls into an absorbing state filled only with prey. In one‑dimensional lattices this transition belongs to the directed percolation universality class, confirming the generic link between single‑absorbing‑state non‑equilibrium models and DP criticality.

Building on this, the authors introduce a multi‑species generalization that incorporates competition for both prey and space. Here a mutation rate μ sets the evolutionary time scale. Low μ produces very long extinction‑speciation cycles, which the authors compare to the ~26‑million‑year periodicity of mass extinctions proposed by Raup and Sepkoski. They suggest that the period of the oscillations in their model is proportional to 1/μ, offering a possible mechanistic explanation for paleontological periodicities. Additionally, by varying a control parameter they observe a transition from a single‑species “replicator” regime to a stable multi‑species ecosystem, which they liken to the hypothesized early‑life transition from a universal code to a diversified biosphere.

The second major part of the paper addresses language emergence using an evolutionary version of the naming‑game model introduced by Steels. Agents engage in pairwise communication events, learning a shared vocabulary through reinforcement. When the density of linguistic interactions exceeds a critical threshold, a Baldwin‑effect‑like coupling of learning and genetic evolution creates a rapid “bio‑linguistic” transition: linguistic abilities become genetically encoded, and the population abruptly reaches linguistic coherence. This demonstrates that language evolution cannot be treated as a purely cultural process; instead, it emerges from the interplay of learning dynamics and Darwinian selection.

In the discussion, the authors emphasize common themes across ecological and linguistic models: multi‑scale dynamics, non‑linear feedback, critical phase transitions, and the importance of incorporating both ecological (short‑term) and evolutionary (long‑term) processes. They advocate for further work that calibrates model parameters with empirical data, explores higher‑dimensional spatial structures, and investigates the robustness of the observed universality classes. Overall, the paper argues that computational modeling provides a unifying framework for studying complex evolutionary phenomena, from species turnover in ecosystems to the rise of human language.