Ten Principles of the Thelen-Barrett-Levin Constructionist Synthesis
Esther Thelen, Lisa Feldman Barrett, and Michael Levin have all done research that can usefully be synthesized under the umbrella of psychological constructionism, for a sufficiently general understanding of “psychological”. There are plenty of other names that should be credited for contributing to the burgeoning constructionist synthesis, many of which are cited in the works of Thelen, Barrett, and Levin. I focus on these three in part because each has produced a body of work that is comprehensible and engrossing to nonspecialists, which is vital to producing a synthesis that spans fields and sciences, and in part because they’re the ones whose work I read the most.
This synthesis generalizes common themes across their work to evince principles that I expect apply to the study of complex systems of any form. Note that this is my generalization, and any subset of Thelen, Barrett, and Levin might disagree with any of these claims.
Principle 1: No commands, no instructions
Many popular scientific theories of the behavior of complex phenomena attribute the behavior to commands or instructions either issued by some governing system or contained within the parts that exhibit or perform the behavior. For example, the behavior of cells and the emergence of various phenotypes is attributed to the instructions contained in DNA; motor behavior is explained by commands issued by the brain or central nervous system; and the behavior of atoms is predicted by their presumed obedience to a set of hypothetical laws of physics. Some scientists try to make room for human entities to have “free will”, where the behavior of people is not wholly reducible to the instructions and commands contained and issued within them; other scientists reject even this exception.
The constructionist synthesis rejects the existence of commands and instructions as empirically unproven and analytically unhelpful. Genetic instructions do not explain motor development; neural commands do not and cannot explain motor behavior; the brain does not act as a central dictator in the body but instead serves as a resource-regulator by issuing predictions about the internal economic efficiency of the body; emotions and other psychological phenomena are not due to evolutionarily designed programs or modules in the brain; so-called reflexive behavior is not obligatory but is instead dynamic problem-solving with respect to the environment; the development of multicellular organisms is not due to instructions found in DNA.
Instead of physics or chemistry, economics provides the general theoretical background structure for the constructionist synthesis. Economics has no terms for commands or instructions, and therefore the behavior of parts and systems cannot be explained or predicted in terms of such nonexistent and indescribable phenomena. For example, suppose that a robber points a gun at you and demands that you give them your money. In ordinary language, we would say that you “don’t have a choice” and that the robber is “in control” and “giving commands”. However, in economics, the robber’s “commands” are just information about the likely consequences of your actions; you still have every ability to make choices in the economic sense of making an optimal selection from your available alternatives. In general, economic agents are always making choices, not being commanded or instructed.
The categorical rejection of commands indicated by the statement of this first principle is a generalization of the empirical and theoretical observations described in the preceding paragraphs. Predictions about commands and instructions have failed in many contexts; and there is no direct evidence for the existence of actual physical laws, only highly reliable empirical patterns. Meanwhile, we know that choice exists in some contexts because people make choices. It would be theoretically awkward to say that the behavior of complex phenomena is sometimes described by commands and instructions and other times by choices. Instead, it is simpler to reject the existence of commands and instructions until and unless evidence arises for their existence, and in the meantime to attempt to describe the behavior of complex phenomena in terms of choices until and unless these predictions fail.
Principle 2: Complex phenomena are constructed/assembled/confabulated
If nothing happens because it is commanded or instructed to happen, then how does anything happen? In the context of psychological phenomena, Lisa Feldman Barrett says that events are constructed. In the context of motor behavior, Esther Thelen says that behaviors are assembled. In the context of memory, Michael Levin says that memories are confabulated. Construction, assembly, and confabulation are three different words for the same thing: putting parts together to make a greater whole.
In the theory of constructed emotion, emotions are constructed out of psychological primitives. To construct happiness, the brain does not have a happiness button it can push in the form of activating a happiness module or happiness circuit. Instead, it builds the emotion out of available parts, i.e., neurons, to fit the sensory situation at hand. In the dynamic systems theory of development, development, whether of a body such as the growth of a human organism or a behavior such as an infant learning to walk, occurs as the evolution of a self-organizing system that spontaneously reorganizes in order to achieve goal states with respect to its environment. In morphogenesis, cells confabulate novel solutions to novel problems—justifying the term “confabulate” because the solution did not previously exist, and is therefore made up by the cell or cells—creating a multicellular organism in the process, such as when cells are able to create consistently sized kidney tubules in newts under varying cell size conditions.
Constructionism occurs in many contexts. Memories are not stored somewhere but are instead constructed as memories. Beliefs about the permanence of objects are put together based on experience. In economics, businesses, industries, and markets are constructed, as are preferences, values, and rationality; you do not have a stored list of your favorite things in your head to call upon whenever you go shopping.
Phenomena are constructed in a goal-directed manner. Emotions are constructed to regulate affect; walking is constructed because an infant wants to move somewhere; multicellular organisms are constructed by the efforts of cells to achieve a shared model of a collective form. Essentially, something is always being constructed because parts are always interacting with each other; goals bias what is constructed. In the social and biological sciences, when something is being constructed that is of interest to sciences, it is goal-directed; this is probably also true for the physical sciences as well.
Principle 3: Categories, not natural kinds
Theories of emotion, motor behavior and development, and morphogenesis have been hindered by scientists theorizing as if they believe in natural kinds, meaning that scientists implicitly, and sometimes explicitly, assume that the categories they lump phenomena into can be distinguished from each other in a perceiver-independent, goal-independent manner. For example, theories of emotion have often assumed that there is such a thing as happiness that is distinct from such a thing as sadness in a way that does not depend on the categorization decisions of the observer but reflects some perspective-independent distinction in some imagined objective reality.
The constructionist synthesis rejects natural kinds. Emotions are not natural kinds; instances of motor behaviors that we may categorize as walking, hitting, etc., are really just particular nodes in a coordinative structure; seemingly distinct phenomena such as life do not have definitions. Relatedly, psychological phenomena in general are all instances of a domain-general process; mental competencies that we are seemingly equipped with such as object permanence are actually learned rather than programmed in. Even in math, natural kinds and definitions do not exist, at least not as people usually imagine them.
When we perceive an instance of happiness, walking, or life, what we are really doing is classifying a set of observations according to categories that are themselves constructed in the moment based on what worked in the past. The brain is forming categories all the time, which it does because categories are used to navigate the world rather than for objective factual accuracy. Categorization decisions emphasize some features of the phenomenon in question while deemphasizing others, creating relevant relationships and exploitable properties based on your goals and your own internal state.
As discussed in later principles, the idea that all classifications of phenomena are subject goal-based decisions rather than objective, perceiver-independent acknowledgments has significant methodological implications for the study of complex phenomena.
Principle 4: Constructed instances are softly assembled
The construction of an instance of a category, such as an emotion, a motor behavior, or a form of an organism, would be greatly constrained if it had to be constructed the same way every time with the same construction materials every time. Instead, constructed entities are formed out of whatever resources happen to be available, a phenomenon known as soft assembly.
In motor behavior, soft assembly occurs at multiple scales. The task of locomotion can be constructed in many ways with many different bodily construction materials. For example, people can walk with their legs, crawl on their hands and knees, drag themselves along the ground with their arms, or even walk on their hands. At a smaller scale, each instance of a motor behavior is softly assembled with different motor neurons that bind to different receptors depending on what is available, and each behavior may be constructed with different physical dynamics regarding exactly which muscles flex and where and how forces are transmitted through tendons, etc. This allows motor behavior to be constructed in a way that accounts for the specific conditions of the body and its current relationship with the environment, such as posture.
The soft assembly of emotion and other psychological phenomena is seen most clearly in the case of degeneracy, where the same psychological phenomenon can be constructed with different parts of the brain, different neural circuits, and different neurons. This allows psychological phenomena to be constructed in a way that accounts for the current constraints on the brain, such as other tasks it is using neural resources to perform. In morphogenesis, soft assembly can be seen in, e.g., the ability to reconstruct memories despite massive brain remodeling, which would not be possible if the memories required dedicated parts.
This principle relates to the principle of no commands: instead of being told what to do, the system itself finds solutions based on available resources and environmental context, allowing for solutions to be flexibly constructed to fit the particulars of the situation, especially the internal situation.
Soft assembly is a familiar idea in economics, where how, eg., houses, are constructed depends on available local resources and conditions. A one-size-fits-all approach would lead to few if any satisfied with their living arrangements while also being much more costly and inflexible in the face of both supply and demand shocks.
Principle 5: The constructing intelligence is collective
Scientists often attribute some competency to a controlling executive or specialized subsystem where the competency is thought to reside. For example, motor behavior may be often attributed to a motor program in the brain, an emotion such as fear may be attributed to a dedicated fear circuit, and morphogenesis may be attributed to some set of genetic instructions. The constructionist synthesis denies that competencies are “anywhere in particular” but are instead constructed out of an ordering of parts, which, in combination with the principle of soft assembly, makes it impossible to attribute competencies to any singular entity or subsystem (including the collective itself, as any part of the collective could be conceivably destroyed or transformed without harming, or at least eliminating, the competency in question).
In motor behavior, the ability to walk is not contained in the brain, nor is it attributable to any specific subsystem of the legs or elsewhere in the body, but is softly assembled from interactions among the parts. Emotions are whole brain-body phenomena; morphogenesis is something cells produce in interaction with each other. In general, there is no Indivisible Whole Part That Does Stuff.
This principle can be generalized to relational realism, the idea that the properties of an object are due to how the object relates to other objects, not to the object itself.
Principle 6: Macro “landscape warping”, micro problem solving
As per principle 5, it is incorrect to attribute the construction of complex behaviors to any particular part or “level”. In general, what humans see as “higher-level” subsystems send signals that warp the “landscape”, such as the energy landscape or activity landscape, of the subsystems that humans see as “lower-level”, which solve problems, or optimize, with respect to this warped landscape. When the lower-level subsystems solve their own problems in this warped landscape, their aggregated behavior ends up solving the system-level problems, without the lower-level subsystems necessarily intending this.
In motor behavior, infants can be induced to construct instances of walking via a treadmill that warps the energy landscape pertaining to the impetus on their legs. Emotions are a class of concepts that refer to the biased landscape for the construction of action and perception, a biasing driven by signals that constitute multimodal summaries from large, densely interconnected neurons. In biology and evolution, organisms can solve new problems by deforming the landscape for their problem-solving parts to construct new solutions.
In economics, this is how the price system controls the economy: prices warp the allocation-and-production landscape for people, who then solve their own problems given the set of prices, which yields collective solutions to problems in allocation space and production space. Without being able to issue commands, this is the only way that higher-level systems can influence lower-level systems to produce collective solutions.
Principle 7: Subsystems are autonomous
Because the problem-solving abilities of complex systems depend on the problem-solving abilities of the parts that make up the system, the properties of the parts cannot be ignored when attempting to understand everything about the properties of the system. Their autonomous behavior is crucial to the construction of the actions and perceptions of the larger system; their preferences, goals, and capabilities cannot be ignored but must instead be predicted, manipulated, and relied upon. This relates to the principle of no commands: If the systems are not being instructed on what to do, then they must be choosing what to do.
In the case of motor behavior, the properties of the legs—how they swing, bend, stretch, etc.—are crucial for understanding how walking is made possible, in particular the resemblance legs bear to simple physical systems such as springs and pendulums. The optimization of the parts in terms of energy and forces is necessary for the optimization of the whole in terms of the system’s goal states. The subsystems that produce motor activity are autonomous, particularly in the sense that if the system-level optimization is inconsistent with their individual goal states, then they will act to change their states and therefore the state of the system. In psychology, interoception, which is the brain’s model of the events occurring inside the body, would be useless if the body was not autonomous; otherwise, its activity would not provide information to the brain. In biology, autonomy at multiple scales is crucial for understanding morphogenesis and evolution.
In economics, the autonomy of the subsystems at multiple scales is evident: people are autonomous, firms are autonomous, and the economy is autonomous.
Principle 8: Multicausality
Scientists often ask questions like, “What’s the mechanism behind [that interesting phenomenon]?” or “Which chemical/molecule/signal/etc. is responsible for that happening?”, etc. The constructionist synthesis says that the implicit assumption of monocausality in these questions is often fallacious. Instead, complex systems construct actions and perceptions in a multicausal way, with no priority given to any particular cause.
In motor development, the infant stepping reflex was historically explained in a single-causal manner in terms of cortical behavior; modern explanations involve muscle flexion, gravity, weight, and bouyancy. More broadly, consider the question of asking what part of the body causes basketball play in basketball players. If you break the left arm of a basketball player, they will stop playing basketball until it heals, which makes it tempting to say that the left arm causes basketball. But the same will also happen if you break their right arm, or their left or right leg, or their back, or their neck, or their fingers or toes, or remove their heart or liver. A single-causal explanation of basketball tries to pinpiont which part is uniquely responsible for basketball; a multicausal explanation tries to understand how the parts interact in a mutually interdependent way to produce basketball play. Multicausality is also seen in the whole brain-body construction of psychological phenomena, in which a confluence of internal factors from every subsystem of the body—immune, endocrine, etc.—determine the resulting construction. The multi-scale architecture responsible for morphogenetic outcomes is also an example of multicausality, with no particular scale having causal primacy.
Economies exhibit multicausality, e.g., economic growth is multicausal.
Principle 9: Complex systems have hidden competencies
Complex systems can exhibit unexpected capabilities in novel environmental conditions, or with changes to their internal structure. The ability of a complex system to accomplish some goal in a given environment cannot be deduced but must be tested empirically.
In developmental psychology, hidden competencies include phenomena as diverse as infant treadmill stepping and the ability of rat pups to ingest semisolid food during a time that they would normally be suckling. In psychology, examples of hidden competencies include the ability of a person with a destroyed amygdala to feel fear and the ability of people with Parkinson’s disease to move quickly when threatened by a horse or a fire. Even algorithms demonstrate hidden competencies; in general, the ability of a system to navigate novel challenges without preprogrammed solutions is crucial to morphogenesis and complex behavior.
This principle can be thought of as a plea for empiricism in evaluating the properties of a system. Scientists are sometimes too quick to assume that they know what a system can do. The only way to be certain is to check.
Hidden competencies in economics includes The Fable of the Bees and The Lighthouse in Economics.
Principle 10: Variation is the signal, not the noise
When studying the behavior simple, static systems, there is often the expectation of a "true” value around which all variation is noise or error, such as the time it takes for a ball of given mass to slide down a slope. Averaging out the noise or otherwise trying to dismiss or minimize the variation is a natural methodological outcome. However, when studying comples, dynamic systems, this method is inappropriate. Quite often, the variation is where all the action is. Indeed, the statistical discovery of modal patterns might be entirely misleading, as such patterns might be more superficial and less informative than the unaltered data.
In motor behavior and development, variability gives the system things to explore and therefore opportunities to find new solutions. Variation isn’t error, it’s testing. In psychology, Lisa Feldman Barrett’s empirical results demonstrate that variation is the norm. She advocates for population thinking, meaning that instances of a concept don’t share some essence but instead are grouped together for a functional purpose; in particular, this demands a radical change in how the role of context is understood in constructing psychological phenomena. Variability is also a crucial fact in morphogenesis; dismissing it as “error” or “noise” would throw a significant set of clues as to how morphogenesis works. In general, the greatest harm of violating this principle is that it gives scientists permission to downgrade, ignore, or discard valuable data.
Economics may be unusual in being relatively tolerant of variability, as we naturally expect people to have different preferences, different skills, etc.