Life and Fitness Landscapes
What is life? How do we know when something is alive? Is fire alive? Is a meme alive? From this incredible process stems all biodiversity, all plants and animals, everyone we have ever known, and ourselves. In this piece we’ll explore a little more about what life is, what it does, and an interesting way to model some of these dynamics.
Cedar branches from Whatcom Falls Park in Bellingham, Washington
What is Life?
Most definitions of life come down to “self-reproduction, with variations”. What does it mean to reproduce? The term “Autopoiesis” was coined to describe this phenomenon, literally meaning “self-creation,” but notice Re-Production is already a very close fit. When entities reproduce, they pass on their genetic material in a fashion which goes on to grow and mature and, if all goes well, propagate again.
We can break reproduction down a little more: life is what Biochemist and Systems Theorist Stuart Kauffman calls an “Autocatalytic” process — a chemical feedback-loop which reacts with its environment in a way that sustains itself. On some level life is nothing but an unfolding (hyper-complicated!) sequence of self-perpetuating chemical reactions.
A microscope photo I took of a flower petal in 2014
But this picture is missing something - what about planning, behavior, and information? A genetic sequence being passed on is at the end of the day a transfer of information which goes on to steer the organization of the organism. In fact, we even know the size of the human genome, “about 3E9 bytes of data”. Where does information fit into this chemical model?
We can think of information as an “emergent phenomenon” which arises out of the collective behavior of a number of parts. “Emergence” is a fuzzy concept for many of us, but we are truly surrounded by examples — one wooden board does nothing to hold up our books, but we can take many boards and nails and construct a bookcase. Where in these materials was the bookcase? How did we create the ability to hold up books out of parts that could not? Similarly, is there a single part of a car allows us to drive around? The concept of Emergence is a rabbit hole and deserves its own essay, but hopefully these challenges are sufficient for now.
We can think of Information as emerging from the relationship of many parts, none of which have information on their own. And, as it turns out, information can reproduce. Life is not just a set of auto-catalytic chemical reactions, but also the use of information (genetic or otherwise) to facilitate these reactions. Artemy Kolchinsky and David Wolpert of the Santa Fe Institute refer to this as “Semantic Information”, information that increases the fitness of a system. Semantic Information is the information responsible for the propagation and fitness of an organism, and, by extension, for catalyzing and propagating more chemical reactions.
Organisms or systems with the right information and organization to propagate in their environment tend to keep propagating (this is known as Criticality). Systems that don’t do that tend not to stick around. In this way life is always evolving, changing and searching for approaches to increase fitness, its ability to sustain and reproduce itself.
Fitness Landscapes
If you are familiar with hill-climbing algorithms, this will probably look familiar - because in a certain sense, that’s what life is - some sort of hill-climbing algorithm towards fitness. In this model, each pip is a new and unique offspring, slightly varied from its predecessors. Those which are more fit within their environments have a higher chance of propagating, effectively “climbing the hill” toward fitness. In this context, the hills correlate to the different “niches” or opportunities presented by the landscape.
Environments and niches are rarely static however - what does it look like when the environment is constantly changing? How does that change the evolution of these organisms?
We can see here that evolution in a dynamic environment is also a dynamic process - the goal for “what to optimize for” is constantly changing — what used to be an optimal strategy can become irrelevant, and what used to be a losing strategy can skyrocket to sudden success. This is a useful way to think about the optimization of systems in general, where the environments of our lives, organizations, and cultures are always shifting and changing.
Notice in the above animation that on flatter parts of the landscape, there is more variation and change in strategy, while in the steep peaks, there is a strong convergence toward one single dominant strategy.
Also notice that over time, in a dynamic landscape, the many entities will tend to converge into a few peaks, or dominant approaches — as the environment changes, the agents will tend to behave more and more similarly. This has implications in decision-making and strategy, where in rapidly changing landscapes, there is more convergence, even if that hill is not a global optimum. In effect this uniformity leads to more systemic fragility, because the agents are exploring a smaller search-space. Imagine the times in your life when you have seen large numbers of people using sub-optimal solutions — in these simulations the dynamics that lead to those outcomes can be seen in a proto-form.
Fitness landscapes are just one way to visualize the evolution and change in our world, and my goal for these labs will be to explore a few more.
