3

I would like to know what is the state of the art.

There are several approaches in epistemology, from the various variants of positivism, to pragmatism, and many others.

In cybernetics a system creates a model of the world and this model is updated based on the information collected by the system. To obtain more information, actions may be performed.

A very simple model (maybe the most simple one) would be the one proposed for artificial curiosity.

I would like to know whether this has been studied and what is the current state of the art wrt the implications of cybernetics to

The type of questions I would like to address are:

  • Could cybernetics be a "better" method than the scientific method? Under which assumptions?
  • Could this be defined as something more general than the scientific method?
  • What kind of knowledge can we get by applying the principles of cybernetics?
  • Which are the limitations of these principles?
  • How is it compared with other methods?

Please remember I am not asking these questions, these are examples of the questions that the references I'm searching for should address.

Thank you.

Trylks
  • 682
  • 5
  • 13
  • 3
    I don't understand what you think the differences are between cybernetics and "the scientific method". Your cybernetic system is allowed to follow any portion of the scientific method, and if a cybernetic system robustly computes a predicted outcome then that is as good a hypothesis/prediction as anything other for testing a theory. – Rex Kerr Jul 29 '13 at 19:05
  • @RexKerr The scientific method adds restrictions about how the experiments should be controlled and repeatable, and many others. A cybernetic system does not add these restrictions, there is simply information in (possibly out as well) and the generation of a model, AFAIK. – Trylks Jul 29 '13 at 19:12
  • Well, if you mean the classical philosphical definition of "the scientific method" as opposed to the one actually used by scientists, maybe so. Forming an evidence-based probability distribution over a family of models is perfectly comfortable for scientists with a strong quantitative background as a natural extension of the "do an experiment to definitively rule out a hypothesis" idea. The standard description is one third human sociology, one third tips on how to make your math easier, and one third the intellectual core of the method. – Rex Kerr Jul 29 '13 at 19:18
  • @RexKerr thank you :) . Although I'm now more confused :/ . Do you know about a comprehensive and brief description of this "new scientific method". I still think the comparison with cybernetics has relevance. For instance, if you see your friend walking into the kitchen you have information about that, but that information is not really scientific, right? This leads to other questions about the relation between epistemology and information theory, but I'm not sure I can write a proper question about that, I'll think about that and try, hopefully in a near future. – Trylks Jul 29 '13 at 19:26
  • 2
    Welllll, this isn't really a "new scientific method". The core is that you have some sort of hypothesis (model), and you test it against data with well-described parameters. It's in practically every non-philosophical description of the scientific method, e.g. Wikipedia: "researchers propose hypotheses as explanations of phenomena, and design experimental studies to test these hypotheses via predictions which can be derived from them. These steps must be repeatable, to guard against mistake or confusion". Experimental design can be as simple as "I look at more stars". – Rex Kerr Jul 29 '13 at 19:44
  • 2
    I guess it would be different if "cybernetics" used models that could not be used to make predictions, but it's hard to make a model that is useful in any way without being able to turn it easily into a predictive engine. – Rex Kerr Jul 29 '13 at 19:47
  • I agree with you that we can consider an scientific observation seeing someone walking into the kitchen, and predict that person will be there for a while and even more stuff depending on more information. However I'm not sure about how many people could agree on us about this broad perspective about science. Do you have any citation in a handy place? Because I think people would agree more with a cooler name than Trylks, e.g. Putnam sounds better. Citing saves a lot of text and time too. Thank you. – Trylks Jul 30 '13 at 17:58
  • 1
    Sorry, I don't have a quote handy. I recall issues like these touched upon even by e.g. Popper, but it's taken modern machine learning and big data science to really focus attention on this type of science as central rather than sort of an odd case to mention in passing and not worry about too much. I'm not up-to-date with modern thinking on the philosophy of science; I've mostly read classics, and even those not recently. – Rex Kerr Jul 30 '13 at 18:38
  • Random thought, no references - AFAIK attempts to use AI for science have a massive weakness in coming up with hypotheses (or guesses as Feynman would call it) that seem to be so easy for humans. Maybe your cybernetic approach to science-like process could do something useful and science-like without the need for this guessing component. – obelia Aug 02 '13 at 17:22
  • @obelia sure, there is no need for hypotheses simply a model is constructed and updated constantly. An hypothesis is simply a part of a model that is specifically tested, in this case the whole model would be tested. – Trylks Aug 02 '13 at 20:56

1 Answers1

1

The thing is that cybernetics as field basically died, mostly because of modern computers. It still has a huge legacy, but it is really hard to say where cybernetics ends and AI, cognitive science, systems biology, information theory and so forth begins. The days when people considered cybernetics to be a new and general science are long past - some ideas have be retained and integrated with modern science, others have been left to stagnate (in particular the abstract, super-general, mathematical formalisations characteristic of their approach).

In recent years various developments in machine learning and Bayesian probability have been identified with cybernetics (most notably, the Bayesian Brain), and these themselves have drawn a lot of attention from epistemologists. Still, the connection is not a very strong one as cybernetics has basically ceased to exist.

Many enactivists have sympathy for cybernetic ideas, looking at their work would probably be a good starting for modern philosophical discussions. The journal "Adaptive Behaviour" would be a good place to look if you are interested in its legacy.

Lucas
  • 1,806
  • 1
  • 12
  • 26