Open-ended evolution – the presence of rich evolutionary dynamics that continuously produce novel, complex communities and species – is a key feature of the natural world. Understanding the conditions that enable open-endedness is a major challenge in artificial life and evolutionary computation. The MODES toolbox, consisting of metrics for detecting change, novelty, ecology, and complexity, is a promising approach for quantifying open-endedness. However, MODES has only been applied to a few systems so far, with limited opportunity for controlled experiments or cross-system comparisons. To address this gap, we implement a custom digital evolution platform (Evo-Sandbox) designed specifically for this purpose. Evo-Sandbox includes configurable modules for that can be combinatorially combined to create diverse environments. We investigate two diversity promoting mechanisms, fit-when-rare, and parasites, to test MODES across a range of conditions. Our experiments reveal that the regions of parameter space in which different hallmarks of open-endedness are maximized are non-intuitive, and that MODES is, in fact, a valuable tool for understanding the resulting behaviors.

When a new evolutionary dynamic is identified, researchers often struggle to understand its long-term effects on evolutionary outcomes. Evolutionary prediction is always challenging, as subtle nuances of dynamics can interact in unpredictable ways. Digital evolution systems, however, provide an empirical alternative to prediction: automated replay experiments can be conducted in large numbers to measure a real distribution of outcomes from a given starting point. Changes in distributions over time can help us understand the long-term implications of seemingly minor events during evolution. We apply this technique to “adaptive momentum”, a new framework that explains how phenomena like selective sweeps can temporarily weaken selection and enhance the likelihood of crossing deleterious fitness valleys. We show that deleterious mutations along the leading edge of a selective sweep can have an outsized influence on the evolutionary fate of a population. Indeed, we see that evolutionary potential to cross new deleterious valleys drastically increases during selective sweeps. Moreover, each valley crossing initiates a new sweep, increasing the potential for further discoveries; this increased potential subsides only once all sweeps have concluded. While we still have much to learn about both adaptive momentum and the role of history in evolution, this work identifies important evolutionary dynamics at play and hones our tools for further studies.

This study investigates the relationship between sparse computation and evolution in various models using a simple function we call sparsify . We use the sparsify function to alter the sparsity of arbitrary matrices during evolutionary search. The sparsify function is tested on a recurrent neural network, a gene interaction matrix, and a gene regulatory network in the context of four different optimization problems. We demonstrate that the function positively affects evolutionary adaptation. Furthermore, this study shows that the sparsify function enables automatic meta-adaptation of sparsity for the discovery of better solutions. Overall, the findings suggest that the sparsify function can be a valuable tool to improve the optimization of complex systems.

During certain evolutionary scenarios, such as genetic sweeps and range expansions, the driving lineages have an increased competitiveness or experience an absence of competition, which results in a higher tolerance of deleterious mutations. We have named this phenomenon, during which individuals have more freedom to explore their fitness landscape, the “Free-for-All” effect (FFA). We present evidence for the free-for-all effect and discuss some of its implications for evolutionary science. This document summarizes work that we are preparing for publication.

The major evolutionary transition to multicellularity shifted the unit of selection from individual cells to multicellular organisms. Constituent cells must regulate their growth and cooperate to benefit the whole organism, even when such behaviors would have been maladaptive were they free living. Mutations that disrupt cellular cooperation can lead to various ailments, including physical deformities and cancer. Organisms therefore employ mechanisms to enforce cooperation, such as error correction, policing, and genetic robustness. We built a simulation to study this last mechanism under a range of evolutionary conditions. Specifically, we asked: How does genetic robustness against cellular cheating evolve in multicellular organisms? We focused on early multicellular organisms (with only one cell type) where cells must control their growth to avoid overwriting each other. In our model, unrestrained cells will outcompete restrained cells within an organism, but restrained cells alone will result in faster reproduction for the organism. Ultimately, we demonstrate a clear selective pressure for genetic robustness in multicellular organisms and show that this pressure increases with the total number of cells in the organism.

Brains are among the most complex evolved objects. In recent years we have seen an explosion in the development of artificial cognitive systems constructed in silico (i.e. digital brains). In fact, we are now capable of creating digital brains whose operation is so complex that they are effectively black boxes (Castelvecchi, 2016; Gunning, 2017). Previous work (Marstaller et al., 2013; Hintze et al., 2018; Kirkpatrick and Hintze, 2019) has identified and expanded upon various information-theoretic measures that can shed light on the internal processes of digital brains. Here we introduce a new information-theoretic measure called Fragmentation ( F ) which can measure how fragmented information is in an a digital brain. To provide a example of the application of F we look at the evolutionary emergence of complexity. Questions regarding the evolution of complexity have been of interest for as long as evolution has been a theory (Gregory, 1935). Nature is responsible for the development of a massive array of complex organisms, each comprised of various organs and regulatory systems that are themselves complex (McShea and Brandon, 2010). It has been observed that complexity can evolve even when complexity itself is being selected against (Beslon et al., 2021). We conclude by using F to show a case of evolved complexity that results in coincidental encryption.

Selection can be described as a filtering process which changes a population over time with regard to the result of some evaluation (i.e. a fitness function). We are interested understanding the relationship between different parameters for altering selection strength and rates of adaptation. In this work we perform a detailed assay exploring the relationship between population size, noisy phenotype evaluation, and tournament size, and their effects on rates of genomic change. We run our model on nearly 4,500 different scenarios. We observe evolution on a smooth fitness landscape as well as nine deceptive landscapes using our model. We show that for the smooth landscape it is always best to have strong selection with noise-free fitness and a large population. For deceptive landscapes, there is an optimum configuration of tournament size and noise that balances exploration and exploitation. Population size, on the other hand, always increases genomic change when larger, because it not only increases selection strength but also maximizes mutational inflow and standing variation. We see that while these parameters for selection strength have similar effects, they each behave in unique ways. Finally, we suggest that evaluation noise is a better proxy for selection strength than population size.

Understanding the structure and evolution of cognition is a topic of broad scientific interest. Computational substrates are ideal for conducting investigations into this topic because they can be incorporated in rapidly evolving Artificial Life systems and are easy to manipulate. However, design differences between currently existing digital systems make it difficult to identify which manipulations are responsible for broad patterns in evolved behavior. This is further confounded if we are trying to disentangle how multiple features interact. Here we systematically analyze components from two evolvable digital neural substrates (Recurrent Artificial Neural Networks (RNNs) and Markov brains) to develop a proof-of-concept for a comparative hybrid approach. We identified elements of the logic and memory storage architectures in each substrate, then altered and recombined properties of the original substrates to create hybrid substrates. In particular, we chose to investigate the differences between RNNs and Markov Brains relating to network sparsity, whether memory is discrete or continuous, and the basic logic operator in each substrate. We then tested the original substrates and the hybrids across a suite of distinct environments with different logic and memory requirements. While we observed trends across all three of the axes that we investigated, we identified discreteness of memory as an especially important determinant of performance across our test conditions. However, the specific effect of discretization varied by environment and whether the associated task relied on information integration. Our results demonstrate that the comparative hybrid approach can identify structural components that enable cognition and facilitate task performance across multiple computational structures.

Sexual selection is a powerful yet poorly understood evolutionary force. Research into sexual selection, whether biological, computational, or mathematical, has tended to take a top-down approach studying complex natural systems. Many simplifying assumptions must be made in order to make these systems tractable, but it is unclear if these simplifications result in a system which still represents natural ecological and evolutionary dynamics. Here, we take a bottom-up approach in which we construct simple computational systems from subsets of biologically plausible components and focus on examining the underlying dynamics resulting from the interactions of those components. We use this method to investigate sexual selection in general and the sexy sons theory in particular. The minimally necessary components are therefore genomes, genome-determined displays and preferences, and a process capable of overseeing parent selection and mating. We demonstrate the efficacy of our approach (i.e we observe the evolution of female preference) and provide support for sexy sons theory, including illustrating the oscillatory behavior that developed in the presence of multiple costly display traits.

As the field of Artificial Life advances and grows, we find ourselves in the midst of an increasingly complex ecosystem of software systems. Each system is developed to address particular research objectives, all unified under the common goal of understanding life. Such an ambitious endeavor begets a variety of algorithmic challenges. Many projects have solved some of these problems for individual systems, but these solutions are rarely portable and often must be re-engineered across systems. Here, we propose a community-driven process of developing standards for representing commonly used types of data across our field. These standards will improve software re-use across research groups and allow for easier comparisons of results generated with different artificial life systems. We began the process of developing data standards with two discussion-driven workshops (one at the 2018 Conference for Artificial Life and the other at the 2018 Congress for the BEACON Center for the Study of Evolution in Action). At each of these workshops, we discussed the vision for Artificial Life data standards, proposed and refined a standard for phylogeny (ancestry tree) data, and solicited feedback from attendees. In addition to proposing a general vision and framework for Artificial Life data standards, we release and discuss version 1.0.0 of the standards. This release includes the phylogeny data standard developed at these workshops and several software resources under development to support our proposed phylogeny standards framework.

A great deal of effort in digital evolution research is invested in developing experimental tools. Because each experiment is different and because the emphasis is on generating results, the tools that are developed are usually not designed to be extendable or multipurpose. Here we present MABE , a modular and reconfigurable digital evolution research tool designed to minimize the time from hypotheses generation to hypotheses testing. MABE provides an accessible framework which seeks to increase collaborations and to facilitate reuse by implementing only features that are common to most experiments, while leaving experimentally dependent details up to the user. MABE was initially released in August 2016 and has since then been used to ask questions related to Evolution, Sexual Selection, Psychology, Cognition, Neuroscience, Cooperation, Spatial Navigation and Computer Science.