At the MSU Digital Evolution Laboratory (Devolab), we perform experimental studies on digital organisms with the twin goals of improving our understanding of how natural evolution works, and applying this knowledge to solving computational, engineering, and biological problems.
While we’re widely interested in the field of digital evolution/artificial life at the Devolab, we have various more specific projects currently being investigated. We’re planning on trying to keep the Projects page up-to-date whenever a new major project is undertaken or an old one wrapped up (hopefully with a publication to share!), I decided to draw your attention to our current work now that everything is up-to-date:
Suicidal altruism is the most extreme form of altruism and therefore studying its evolution sheds light on many altruistic behaviors. We previously studied the environmental factors that influence the evolution of suicidal altruism and how quorum sensing influences that behavior. We are currently studying suicidal altruism when the altruistic act produces a public good.
Division of Labor
Cells in your body, leafcutter ants in a eusocial colony, and factory workers in a plant are just a few examples of groups exhibiting division of labor strategies in order to survive and thrive. This research explores hypotheses for why division of labor is an adaptive strategy and the role division of labor plays within the major transitions of evolution. These issues are extremely challenging to study using organic systems because of the slow pace of evolution and imperfect historical data. To address these challenges, we perform experiments using digital organisms (populations of self-replicating computer programs that undergo open-ended evolution). Insights from these experiments shed light on evolutionary questions surrounding division of labor and can also be applied to evolving solutions to engineering problems, such as developing teams of specialist robots that must cooperate to achieve an overall objective.
Major transitions in evolution occur when formerly distinct individuals form a higher-level unit that functions as a single reproductive entity. These transitions can be fraternal, where genetically similar individuals (i.e., close kin) differentiate to perform various tasks, or egalitarian in which formerly distinct organisms create a super-organism that replicates all of its genetic material. A fundamental aspect of major transitions in evolution is the role of division of labor, where lower-level individuals specialize and cooperate as part of a higher-level unit to survive. These transitions raise evolutionary questions regarding the conditions under which formerly distinct individuals would cooperate with others, and once they did, how this arrangement persisted. Such questions are incredibly challenging to study with organic systems due to imperfections in the historical data and long generation times that preclude systematic study within a reasonable time frame. For this project, we use digital evolution, a form of experimental evolution where organisms are self-replicating computer programs, to study questions surrounding major transitions in evolution.
Resource Spatial Heterogeneity
A central question in ecology has long been how such a wide diversity of life can evolve within a seemingly limited set of niches. Many suspect that spatial heterogeneity is an important factor in making this possible, but this is a very challenging principle to test in biological populations. To try and understand whether spatial heterogeneity is a general driver of diversity, we are experimentally testing it in Avida.
Inferring Rugged Adaptive Landscapes via Spatial Structure
Adaptive landscapes are a metaphor for understanding how populations evolve over many generations. Rugged landscapes are problematic for evolution as populations have difficulty reaching optimal genotypes. We are determining the degree of ruggedness of bacteria evolving in a novel environment (using spatial structure) and hope to generalize this method to other laboratory populations.
Traversing Adaptive Landscapes Across Changing Environments
For most (really all) evolving organisms, their environment changes over time, exposing them to different selective pressures. Sometimes, combinations of environments can be useful in directing evolution toward more profitable outcomes. We are investigating the relationship between the organism’s effect and response toward environmental change, especially with regard to achieving more optimal solutions.
Early Evolution of Associative Memory
We have been using the Avida platform to evolve agents capable of navigating a twisting path, signaled with arbitrary cues. In order to succeed at the task, the agents must be able to associate the arbitrary cues with their meanings, such as ‘turn right’ or ‘turn left’, early on and use that knowledge to navigate the remainder of the path.
Using Evolution to Solve Other Problems:
Evolutionary Breast Cancer Detection
Determining whether or not a mammogram image depicts breast cancer is a challenging problem, even for experienced radiologists. We are attempting to evolve agents, controlled by Markov Brains, that explore images to determine whether or not cancer is present in them. We hope that this work will enable us to detect cancer earlier and more accurately.
Anomaly detection and classification
Wireless sensor networks are now being used to collect vast quantities of data. We are developing an algorithm, inspired by robotics algorithms, to facilitate the processing of this data. Our approach is to categorize time points and then leverage relationships between variables within these categories to detect anomalous events and classify them as either the result of sensor faults or non-erroneous rare events. We hope that this will ultimately prove to be a useful tool, both for maintaining and utilizing sensor networks, and for enabling them to make smarter decisions about what data to collect.
A New Digital Evolution Platform Based on Gene Regulation Dynamics
We want to use digital evolution to study cells’ regulatory networks. We believe this can help us understand the mechanisms behind behavioral decisions of the simplest organisms. This project will require extensive software development, and we hope it will result in a platform that can reproduce many of the regulatory circuits observed in real organisms.
As you can see, we’re working on a lot of cool projects! I’ll be sure to update you when the Projects page is changed, but also check there for contact information for each of these projects if you’re interested in chatting further.
Hello, world, we’re back! The Devolab (i.e. Dr. Charles Ofria’s research lab) has been up to lots of research, but not much blogging – until now. We’ve got lots of plans for discussing interesting papers, ongoing research, behind-the-scenes looks at grad student life, and much more. For now, plan on seeing new posts on Tuesdays.
To start things off, we thought you’d like to know what we’ve been up to in the past year or so, therefore here is a list of the recent publications from the lab. Going forward, we’ll aim to have posts with high-level discussions of our new papers, and slowly fill in with posts about some of the more exciting previous papers. Without further ado, we present research:
- Heather J. Goldsby, David B. Knoester, Benjamin Kerr, and Charles Ofria. The Effect of Conflicting Pressures on The Evolution of Division of Labor, PLoS One, 2014. link
- Anya E. Johnson, Heather J. Goldsby, Sherri Goings, and Charles Ofria. The Evolution of Kin Inclusivity Levels, GECCO, 2014. link
- Anya E. Johnson, Eli Strauss, Rodney Pickett, Christoph Adami, Ian Dworkin, and Heather J. Goldsby. More Bang for Your Buck: Quorum-Sensing Capabilities Improve the Efficacy of Suicidal Altruism, ALIFE 2014. link extended
- Heather J. Goldsby, David B. Knoester, Charles Ofria, and Benjamin Kerr. The Evolutionary Origin of Somatic Cells Under the Dirty Work Hypothesis, PLoS Biology, 2014. link
- David M. Bryson, Aaron P. Wagner, and Charles Ofria.There and back again: gene-processing hardware for the evolution and robotic deployment of robust navigation strategies, GECCO, 2014. link
- Aaron P. Wagner, Luis Zaman, Ian Dworkin, and Charles Ofria. Behavioral Strategy Chases Promote the Evolution of Prey Intelligence, arXiv, 2014. link
- David B. Knoester, Heather J. Goldsby, and Philip K. McKinley. Genetic Variation and the Evolution of Consensus in Digital Organisms, IEEE Transactions on Evolutionary Computation, 2013. link
- David M. Bryson and Charles Ofria.Understanding Evolutionary Potential in Virtual CPU Instruction Set Architectures, PLoS One, 2013. link
- Arthur W. Covert, Richard E. Lenski, Claus O. Wilke, and Charles Ofria. Experiments on the role of deleterious mutations as stepping stones in adaptive evolution, PNAS, 2013. link
- Laura M. Grabowski, David M. Bryson, Fred C. Dyer, Robert T. Pennock, and Charles Ofria. A Case Study of the De Novo Evolution of a Complex Odometric Behavior in Digital Organisms, PLoS One, 2013. link
- Miguel A. Fortuna, Luis Zaman, Aaron P. Wagner, and Charles Ofria. Evolving Digital Ecological Networks, PLoS Computational Biology, 2013. link
- Christopher H. Chandler, Charles Ofria, and Ian Dworkin. Runaway Sexual Selection Leads to Good Genes, Evolution, 2013. link
- Heather J. Goldsby, Anna Dornhaus, Benjamin Kerr, and Charles Ofria. Task-Switching Costs Promote the Evolution of Division of Labor and Shifts in Individuality, Proceedings of the National Academy of Science, 2012. link
I tried to make sure that all the paper links are open-access, so let me know if you’d like to read a paper that you can’t access above. As always, we’d love to discuss all of this research with you, so let us know if there is a particular one you’d like a more in-depth post on.
Several new Devolab publications have been accepted to appear at Artificial Life 13.
“Digital Evolution Exhibits Surprising Robustness to Poor Design Decisions” by David M. Bryson and Charles Ofria
When designing an evolving software system, a researcher must set many aspects of the representation and inevitably make arbitrary decisions. Here we explore the consequences of poor design decisions in the development of a virtual instruction set in digital evolution systems. We evaluate the introduction of three different severities of poor choices. (1) functionally neutral instructions that water down mutational options, (2) actively deleterious instructions, and (3) a lethal die instruction. We further examine the impact of a high level of neutral bloat on the short term evolutionary potential of genotypes experiencing environmental change. We observed surprising robustness to these poor design decisions across all seven environments designed to analyze a wide range challenges. Analysis of the short term evolutionary potential of genotypes from the principal line of descent of case study populations demonstrated that the negative effects of neutral bloat in a static environment are compensated by retention of evolutionary potential during environmental change.
“Finger-painting Fitness Landscapes: An Interactive Tool for Exploring Complex Evolutionary Dynamics.” by Luis Zaman, Charles Ofria, and Richard E. Lenski
Evolution involves only a few simple processes, yet the resulting dynamics are surprisingly rich and complex. Sewall Wright developed the metaphor of fitness landscapes to provide deeper insight into the complex workings of evolution. Here we extend that metaphor by visualizing in real time the dynamic processes that drive evolution. We allow viewers to construct fitness landscapes interactively while also varying key parameters including population size, mutation effect size, mode of reproduction (asexual or sexual), and densitydependent selection. This application is both mechanistic and visual, and it thereby allows the active exploration of evolutionary processes. We walk the reader through several exercises including both simple activities potentially suitable for education and examples of deeply conceptual topics that remain the focus of current research in evolutionary biology.
“Evolutionary Potential is Maximized at Intermediate Diversity Levels” by Bess L. Walker and Charles Ofria
Diversity in a population is often cited as a major facilitator for the evolution of new complex features. The intuition behind this dynamic is that if a population is exploring multiple regions of a fitness landscape, more opportunities exist to find new functionality. We use the digital evolution software platform Avida to explore the effect of multiple limited resources on phenotypic Shannon diversity and, in turn, on evolvability of populations. We show that Shannon diversity peaks at intermediate levels of resource availability to the population, and we map the evolvability of a complex computational task on this availability-diversity gradient. While the evolvability of the complex task is highest at intermediate availabilities, it does not peak at the same resource inflow level as Shannon diversity, and it is more robust than diversity in its response to inflow level. These results indicate that while phenotypic Shannon diversity may play into the evolution of complex features, the selective pressures caused by diversity cannot be the only — or indeed even the main — pressures behind such evolution.