Assessing where and how information is stored in biological networks (such as neuronal and genetic networks) is a central task both in neuroscience and in molecular genetics, but most available tools focus on the network’s structure as opposed to its function. Here, we introduce a new information-theoretic tool—information fragmentation analysis—that, given full phenotypic data, allows us to localize information in complex networks, determine how fragmented (across multiple nodes of the network) the information is, and assess the level of encryption of that information. Using information fragmentation matrices we can also create information flow graphs that illustrate how information propagates through these networks. We illustrate the use of this tool by analyzing how artificial brains that evolved in silico solve particular tasks, and show how information fragmentation analysis provides deeper insights into how these brains process information and “think”. The measures of information fragmentation and encryption that result from our methods also quantify complexity of information processing in these networks and how this processing complexity differs between primary exposure to sensory data (early in the lifetime) and later routine processing.

Deep learning (primarily using backpropagation) and neuroevolution are the preeminent methods of optimizing artificial neural networks. However, they often create black boxes that are as hard to understand as the natural brains they seek to mimic. Previous work has identified an information-theoretic tool, referred to as $R$⁠, which allows us to quantify and identify mental representations in artificial cognitive systems. The use of such measures has allowed us to make previous black boxes more transparent. Here we extend $R$ to not only identify where complex computational systems store memory about their environment but also to differentiate between different time points in the past. We show how this extended measure can identify the location of memory related to past experiences in neural networks optimized by deep learning as well as a genetic algorithm.

Group hunting is common among social carnivores, and mechanisms that promote this behavior are a central topic in evolutionary biology. Increased prey capture success and decreased losses from competitors are often invoked as factors promoting group hunting. However, many animal societies have linear dominance hierarchies where access to critical resources is determined by social rank, and group-hunting rewards are shared unequally. Despite this inequality, animals in such societies cooperate to hunt and defend resources. Game theoretic models predict that rank and relative rewards from group hunting vs. solitary hunting affect which hunting strategies will evolve. These predictions are partially supported by empirical work, but data needed to test these predictions are difficult to obtain in natural systems. We use digital evolution to test how social rank and tolerance by dominants of subordinates feeding while sharing spoils from group hunting influence which hunting strategies evolve in digital organisms. We created a computer-simulated world to reflect social and hunting dynamics of spotted hyenas (Crocuta crocuta). We found that group hunting increased as tolerance increased and as the relative payoff from group hunting increased. Also, top-ranking agents were more likely to group hunt than lower-ranking agents under despotic sharing conditions. These results provide insights into mechanisms that may promote cooperation in animal societies structured by dominance hierarchies.

Degeneracy—the ability of different structures to perform the same function—is a fundamental feature of biological systems, contributing to their robustness and evolvability. However, the ubiquity of degeneracy in systems generated through adaptive processes complicates our understanding of the behavioral and computational strategies they employ. In this study, we investigated degeneracy in simple computational agents, known as Markov brains, trained using an artificial evolution algorithm to solve a spatial navigation task with or without associative memory. We analyzed degeneracy at three levels: behavioral, structural, and computational, with a focus on the last. Using information-theoretical concepts, Tononi et al. (1999) proposed a functional measure of degeneracy within biological networks. Here, we extended this approach to compare degeneracy across multiple networks. Using information-theoretical tools and causal analysis, we explored the computational strategies of the evolved agents and quantified their computational degeneracy. Our findings reveal a hierarchy of degenerate solutions, from varied behaviors to diverse structures and computations. Even agents with identical evolved behaviors demonstrated different underlying structures and computations. These results underscore the pervasive nature of degeneracy in neural networks, blurring the lines be- tween the algorithmic and implementation levels in adaptive systems, and highlight the importance of advanced analytical tools to understand their complex behaviors.

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.

: Evolutionary theory seeks to explain the remarkable diversity and adaptability of life on Earth. Current theory offers substantial explanatory power, but it overlooks important transient dynamics that are prominent only when populations are outside equilibrium, such as during selective sweeps. We identify a dynamic that we call “adaptive momentum” whereby lineages with a selective advantage can temporarily sustain more deleterious mutations. This reduction in the strength of purifying selection allows populations to explore fitness valleys that are usually too costly to enter, potentially leading to the discovery of otherwise inaccessible fitness peaks. Using mathematical and agent-based simulations, we demonstrate adaptive momentum and show how periods of disequilibrium become windows of enhanced adaptation. Genetic exploration can occur during these windows without requiring mechanisms such as changing environments or complex landscapes. Adaptive momentum provides a simple potential explanation for bursts of rapid evolution observed in nature, including in pathogens such as SARS-CoV-2 and cancers. (152 words)