Collective intelligence theory as a generalization of dynamic systems theory
One of the first things I noticed about the work of Mike Levin and his lab is the similarities to the work of Esther Thelen and her collaborators. I built up connections between their work and also Lisa Feldman Barrett’s work into something I call the constructionist synthesis, which describes how behavior, development, and thought are all manifested by the process of allostasis, or internal coordination via interoceptive signals.
Esther Thelen worked in dynamic systems theory, which treated developing and behaving creatures as dynamic systems that tend toward attractor states despite obstacles and perturbations. Dynamic systems theory led to some really cool experiments, culminating in her work on the A-not-B experiment, which to this day is the most profound set of experiments on human cognition that I know of. And dynamic systems theory was sufficient for Esther Thelen and her coauthor, Linda Smith, to see the connection between motor behavior and embryogenesis. Additionally, the idea of multicausality that emerged from dynamic systems theory is still very underrated and underutilized, and may prove to be most enduring element of her work.
But there’s no denying that dynamic systems theory eventually ran out of steam and failed to achieve what it hoped. Writing down a bunch of differential equations and having a computer work through them did not prove to be a helpful way to study a developing, behaving, thinking creature. And in some ways the focus on the ideas of dynamic systems distracted from what could have been a more significant and sustained study of the intelligent problem-solving of the developing organism.
There are a few mistakes I think dynamic systems theory made. One mistake was to overemphasize the role of physics in producing motor behavior. The fact that limbs can have spring-like and pendulum-like properties is important for understanding how motor behavior is constructed. However, these ideas just don’t take you very far: most motor behavior is far more varied and adaptive than treadmill walking is, and the stuff about springs and pendulums only does so much to help make sense of the observations.
Another mistake is an overemphasis on the mathematics of dynamic systems. The equations are descriptive, not controlling, and writing them down does not necessarily give you much information about what is going to happen next, let alone why. Dynamic systems may be more fruitfully studied as resources in a “Platonic” space that are searched for and exploited by physical systems as useful patterns in a manner that is not essentially different from an animal searching for food in its environment.
The biggest flaw in dynamic systems theory was simply the failure to become collective intelligence theory. M.T. Turvey came very close to describing motor behavior as an example of collective intelligence. Dynamic systems theory failed to pay sufficient attention to the most important elements of constructing motor behavior: to the idea of subsystems of motor behavior as autonomous systems, i.e., as agents in their own right, and to the idea of functional linkages between subsystems of motor behavior as cognitive glues.
I think that collective intelligence theory can be thought of as a generalization of dynamic systems theory. There are many important similarities between the two theories, but collective intelligence theory successfully broadens some of the ideas lurking in the background of dynamic systems theory. The most striking similarities are:
Development and behavior without instructions or commands: Neither morphogenesis nor motor and cognitive development occurs due to instructions given from the genes. Nor is motor behavior the result of neural commands.
Autonomous parts: Development and behavior both occur because the parts that construct them—the cells, body segments, whatever else—have their own attractor states, as dynamic system theory calls them, or goal states, as collective intelligence theory calls it. Collective intelligence theory advances on dynamic systems theory by viewing the parts of the collective as fully agents in their own right, pursuing their own goals in a self-interested manner.
Linkages: motor behavior occurs in a coordinated fashion because there are linkages between the parts of motor behavior, such as the sharing of tension between body segments, that turn individual parts into functional units. Collective intelligence theory generalizes the idea of linkages to the idea of cognitive glue that binds individuals into a collective whole.
Normal outcomes despite obstacles, perturbations, and novel problems: development and behavior are able to achieve typical outcomes in a regular, reliable, and robust fashion despite unusual and challenging circumstances. Collective intelligence theory improves upon dynamic systems theory by understanding that the intelligence of a system can be studied by seeing how it reaches target states despite difficult and unexpected circumstances.
Hidden skills/hidden competencies: Both dynamic systems theory and collective intelligence theory have produced striking examples of systems demonstrating abilities that no one could have anticipated and which evolution could not have crafted. For example, babies can walk on treadmills well before they begin walking on their own, and cells can negotiate for position with each other to achieve morphogenetic outcomes as if they are economic agents.
Relational realism: I think it’s very striking that dynamic systems theory produced the idea of multicausality, while collective intelligence theory works with the idea of polycomputing. Both ideas are about how we view what a system is and does is about the goal-directed perspectives we take rather than some true essence of the system. Things are what they are in relationship to other things.
Ultimately, dynamic systems theory was a fruitful but limited paradigm. Collective intelligence theory successfully generalizes it by emphasizing the agentic nature of the parts that construct development, behavior, and cognition, and how these parts are coordinated through media that share interoceptive signals across the system. My goal in showing the connections between dynamic systems theory and collective intelligence theory should help to bridge the gap between research in morphogenesis and research in motor behavior. Motor behavior is fast morphogenesis, so the study of development should be able to borrow tremendously from the study of behavior and vice versa.