Or when life gives you X make f(X)
A central question in the biological sciences is “what is life?”. What are the properties that make some system qualify as living and others as non-living?
The hardest scientific challenges are often the ones that seem intuitively easy. We don’t know what makes a system a living one, but we know it when we see it.
What is clear is that life is complex. Living systems are without exception composed of many interacting parts that are interdependent and mutually-constitutive. This is essentially a formulation of one theory of life put forward by biologists Humberto Maturana and Francisco Varela and expanded upon by Evan Thompson termed ‘autopoiesis’. For them, a living system is one organized as a closed network of ‘enabling dependencies’. (They also claim this organizational property is prior to features such as self-replication and heredity in defining life. I’ll avoid that debate for now, but it is worth consideration.)
The leaves and the roots of the plants enable one another’s continued existence. They are also enabled by the sun, but nothing in the plant enables the sun, so it is outside this closed enabling-network and therefore not part of the autopoietic (living) system. However, the sun and the rain are crucial for the plant. All far-from-equilibrium systems need flows of matter and energy to maintain their structure. And life is nothing if not far-from-equilibrium.
The problem I see with autopoiesis as a definition of life is that phenomena we would intuitively identify is non-living seem to fit the bill. Hurricanes have their form maintained over time by virtue of their internal organization (the ‘spiral’ dynamic of the relevant atmosphere is self-reinforcing) coupled with the enabling flow of matter and energy from the warm, wet ocean. This sounds a lot like the plant.*
But a hurricane is not alive. There is a missing ingredient.
In Nassim Taleb’s book Antifragile he drives home the essential distinction between a random variable X and an agent’s exposure to a random variable f(). If we can properly craft f(), we can do a lot less work in trying to predict or control X, and let our payoffs come from f(X) — not the naked X. Taleb goes into great specificity on the form(s) f() might take, and what that means for the system’s gains and losses in the face of uncertainty, but we won’t delve into that here.
For purposes of thinking about the question “what is life?”, we can think of the environment at large as a random (high-dimensional) variable that a system is exposed to and interfaces with.
And here is the key difference: the hurricane is exposed, nakedly, to X. If the water warms => persistence. If a cold front moves into the area => disintegration. The internal structure of the hurricane does not in any way adjust itself to a more favorable exposure for increasing persistence.
But life does. The plant doesn’t merely receive sunlight, it reaches for it. The internally generated behaviors of the plant actively craft their exposure: increasing exposure to enabling dependencies, and decreasing exposure to sources of harm.
This is a non-trivial difference. We summarize with the following speculation:
Living systems are those which are both autopoietic and whose internally generated structures, dynamics, behaviors, and forces are involved in crafting its exposure f(X) to increase its persistence.
*[Note added after initial publication: In correspondence, Evan Thompson has disputed that a hurricane is autopoietic. I am not convinced by his arguments here, as I read them to depend on particular substantial properties rather than organizational properties. Nonetheless, Evan admits of autopoietic systems that are indeed not living, so the dispute over whether hurricanes are autopoietic or not is immaterial to the main thrust of this post.]