© 2012 Joost van der Leij
Hyötyniemi (2006) in Neocybernetics in Biological Systems is advocating the use of multivariate statistical mathematics to study complex systems that so far have defied analysis due to their complexity.1 Hyötyniemi proposes a neocybernetic approach where cybernetics is updated with the idea of emergence. This is a bold and necessary move, one that is a first start to begin to tackle new problems that are finally showing up on our horizon. Like, for instance, the need within artificial intelligence research for computers and robots that program or build themselves for so far any ascribed sign of intelligence is only a compliment to the builder or programmer. It has been suggested that a neocybernetic approach with its important ideas about emergence is a promising avenue for further research.2 So any work on neocybernetics is praiseworthy. At the same time it also makes it more important to be critical.
Neocybernetics in Biological Systems really is an attempt to show what wonderful things we might do once we take the neocybernetic path and start using multivariate statistical mathematics. It tries to persuade us to see the light and persuades us to spend more time and research on neocybernetics. Why is persuasion needed? No doubt the mathematics stands on its own and is solid and true. The real questions here are why would we want to work with simplified, highly abstract models and how do these neocybernetic models connect with reality. Hyötyniemi is looking for a justification for this project and feels that he is justified by providing a philosophical framework for neocybernetics:
The aim of this paper is to discuss three major flaws in this line of reasoning. These flaws have no impact on the main thrust of neocybernetics. They are problematic for its justification though. In short they are the use of frequentist statistics rather than Bayesian statistics, the essentialist nature of the philosophical framework and the use of semantics. All three flaws can be repaired. Repairing the argument in this way will strengthen the case for neocybernetics considerably because in doing so it will also solve many minor objections that crept up as a result of using frequentism, essentialism and semantics. The strategy will be to first show the good that neocybernetics might accomplish. Then to show the two major weaknesses in the approach given by Hyötyniemi. And finally it will be shown how these weaknesses can be overcome and repaired and how this ultimately strengthens the case for neocybernetics.
The term “cybernetics” has been coined by Wiener (1948).4 It has been popularized in modern terms like “cyberspace” and “cyborgs”. it is the culmination of a research program to find ways of computing stuff that is hard to do for humans, like calculating the path of an enemy warplane that needs to be shot down. Wiener’s approach was more organic than Turing’s approach who is the godfather of the digital general purpose computer. A good example of an organic approach to computing is the Cockroach-Controlled Robot, where you have a robot that moves through a room like many digital robots but all the calculations are done by a cockroach.5 Even though cybernetics “lost” the battle for the computer and almost everyone is nowadays working the Turing’s digital computer, cybernetics has been such a success that it has given birth to the field of cognitive science. But this success has a flip side. With the rise of cognitive science the focus on control and communications has been decreased in favor of a focus on cognition. Of course cognition plays an important role within control and communication, but putting cognition before control and communication steers the whole research project into a completely different direction. A direction in which cognitive psychology plays a much, much bigger role and one that, in retrospect, might have been less favorable than originally imagined. With the lack of any landmark breakthroughs the appeal of cognitive science starts to waver ever so little and the first cries for a revitalized cybernetics, neocybernetics, can be heard.
What is needed to advance the study and understanding of complex systems. According to Hyötyniemi General Systems Theory “can easily become too holistic without concrete grounding”6 and computationalism might have chaotic iterations without any correspondence with reality and where small changes in initial settings of algorithms lead to wildly different results. What is needed is not a new science but a new interpretation of established mathematics. Neocybernetics views complex systems in terms of feedback and emergence. Elementary particles give rise to individual atoms. Groups of atoms form and out of them emerge macroscopic entities that can be studied in large volumes where there will be attributes or properties not found on a previous level nor can there be a reduction of these properties or attributes to a lower level. Temperature cannot be described in terms of elementary particles.
Neocybernetics has four key ideas. First is the idea of dynamic balance where complex systems have attractors that create stability within a complex system. This stability is dynamic as the complex system has elasticity and tension so that it moves between different stable states. The second idea is that of environment-orientedness as no cybernetic system can exist in isolation. The third idea is high dimensionality where structural complexity is replaced by dimensional complexity. The final and fourth idea is simplicity where simple complex systems can be used as an analogy for complex complex systems.
Out of these four ideas we get a bunch of very nice emerging results. Here I’ll give four examples but Hyötyniemi has many more. For one you can show how complex systems can become goal-seeking by showing that:
Once you have goal-seeking in a system that fact alone allows us to move from individuals to populations:
“The key point is that following the neocybernetic model there is evolutionary advantage. It turns out that optimality in terms of resource usage is reached, meaning that surviving, successfully competing natural populations assumedly must have adopted this strategy.”8
And once you have populations you can turn from resources to information:
“It is assumed that in a long run an evolutionary surviving system exploits all information it can see: being capable of efficiently exploiting the resources is a prerequisite of surviving in an environment, successful systems are the most active in acquiring for [sic?] more and more information.”9
Then if we are able to show how pure information ties back into biological systems as it must do somehow, we get wonderful stuff like:
“When signals are not existing purely in the infosphere but also, for example, in the chemosphere, a tight web of connections to the environment is constructed, constituting a grounding of “self”. If associations become reconnected, the contents of the feelings can also change – neuro-linguistic programming (NLP) can truly change the way we see the world.”10
Being able to have one general cybernetic description that begins at the smallest level like subcelluar processes of neurons for instance and is able to move from there from the individual to populations and back again to the impact a population makes on the individual is wonderful indeed. So there is every reason to justify a research program into neocybernetics. Unfortunately, the justification as given by Hyötyniemi runs into three major problems: the use of frequentism, essentialism and semantics.