Epistemology, perception and reality

Epistemology, or the theory of the origin and limits of knowledge, comes up quite often on CSGnet, and it looks like there are many different opinions. I’ve found that expressing the variables ans functions in mathematical formulas and equivalent diagrams tends to clear up things for me. Another super valuable trick is from physics - always using the units of variables, and specifying what kind of transformations of units any of the functions is doing. Very, very helpful.

From the thread about diagrams:

I think there is a vagueness here - whos perception are we taking about? And which reality? I think there is only one perceptual experience we can talk about, our own. In that sense, “everything is perception”, and we don’t have any direct knowledge of the reality outside of us, assuming it exists. So, if we cannot measure a variable from outside reality, then we cannot say anything about the correlation between a known, measured perceptual variable and an unknown variable from reality. Correlation being a specific mathematical procedure, like Pearson’s coefficient. We can’t know any sort of dependence if we can’t measure one of the variables.

On the other hand, if we are talking about a perceptual signal, measured in some neural units, and some environmental variable, like a position of an object in meters, then we are necessarily talking about a our own experience as researchers or observers of another control system. We are not talking about some outside reality, but in this case we can certainly find correlations between an environmental and a perceptual variable.

(diagram from Feedback, Beyond Behaviorism; Powers, 1973)

Here is an attempt to put some “reality” into the diagram of a tracking task.

This is a diagram of a model of a feedback loop. The model describes the measurable cursor position (numerical value in pixels), the target position (also pixels) stored in computer memory, and postulates the existence of three neural quantities, the perceptual, reference and error signals, measured in some neural units, maybe rates of firing, population codes or something along those lines, somewhere in the brain of the participant, maybe in the visual system.

The diagram says nothing about the mind of the participant or about his experience. We can assume that the participant has the same visual experience as we do, since we are both human, have similar brains, etc, but there is not need for that assumption. We could also hypothesize that a thermostat has the experience of hot and cold, as we do. But I don’t see any way of proving that the thermostat or our participant has experience, so why bother?

Two things that maybe look like “reality” are the light waves going between the monitor and the eyes of the participant. The visual system creates the perceptual signal from raw light waves. For the visual system, there are is nothing outside but light waves. They are placed in a somewhat arbitrary position. There are also light waves inside the eyes, around the hand the pen, inside the computer. There are forces of different kinds all around the body and the computer too .I’ve put them there to emphasize that on the interfaces between the person and environment, in this case, there are only raw light waves and raw forces, and those are not “outside reality”, they are models we use to explain our observations of colors and movements.

That diagram does not have a controlled variable.
Here is how to get it:

By replacing the system on the left that is taking as input the cursor and target positions with an equivalent system on the right.

We can also replace the error signal to cursor system with an equivalent output function:

The controlled variable is not in the “outside reality”. It is not “in the perception of the participant”. It is a part of the observer’s model of the feedback loop, in the environment part, and can be measured or calculated by finding qc = G(target) + F(cursor). The input function is merely transforming the controlled quantity into the perceptual quantity, just a unit converter, with maybe a delay.

We can say that the variables of cursor and target position are reliably present in the environment, but this is just the environment part of the model designed by the observer, they are his experience. There is no outside reality here.

Yes, there are at least two opinions. One that we do not need any epistemology because everything is just perception and we unproblematically always know our perceptions, and the other that at least sometimes there can be problems of knowing or problems of truth and because of that we need some kind of epistemology. I, for one, belong to the latter group and think that the first alternative is too good for being true. If there were only perceptions and we always knew our perceptions, then no-one were ever wrong, mistaken, surprised about anything. There were never need to check whether you really saw, heard or otherwise perceived something right etc. For example if my wife says me something I must often ask again what she said because I do not hear very well and if I get it wrongly what she said the consequences can be more or less fatal…

But, as for the PCT diagrams, it is important to choose carefully what elements are needed and necessary. If the purpose of the diagram is to model mathematical dependencies between variables then it is clearer to leave reality and epistemological questions out, just like Adam describes. We have no way to determine the mathematical properties of reality, other than perceptions. One would say that epistemology belongs rather to the meta side of the science, but from there it affects how we understand the diagrams and what kind of broader inferences we can make.

According to my current epistemology there is something we can know about reality and about the relationship between perception and reality. (This opinion can be justified for example starting from the idea of the evolution of living beings, like Martin has often stated.) We cannot know the exact mathematical character of that relationship but we can assume that it is – at least often – an analogical dependency function. Because the character of that function remains necessarily fuzzy and/or qualitative for us it has been very traditionally referred to with the concept of truth: In relation to the reality the perception can either true or untrue (or perhaps something between them). Especially in the lowest hierarchical levels it is natural to presume that in normal conditions the perceptions are almost always true. That is just the reason why it is usually unproblematic and useful to leave reality and epistemological references out in typical PCT diagrams of empirical test situations.

I am afraid that in more general and higher-level contexts – than the exact results of tracking tests etc. – which may contain more interesting questions to the broad audience, the claims that there is only perceptions and we need not care about reality make us seem either postmodern or ridiculous or probably both.

Yes! When I was working with Bill Powers on developing various simulations, we found it critically important to do a “unit analysis” to make sure that the transformations involved in the various functions and variables were commensurate. For example, when integrating over time, a velocity in meters/second should yield a number scaled in meters ((meters/second) * seconds = meters). If it doesn’t, you’ve done something wrong. Although such a mistake is unlikely in a simple transformation like that, in more complex transformations such as those involving, say, the computation of Coriolis and other forces involved in limb movement, can easily go astray. A unit check can be very helpful there to assure that the computations make sense.

1 Like

The first paragraph refers to our qualitative subjective experience of perceptions. This is all we directly have.

The second refers to correlations between such perceptions and other perceptions called measurements. The measurement-perceptions and the correlations among them are called data. The taking of measurements and the finding or fashioning of correlations are fundamental disciplines of science. A fundamental purpose of science is to warrant epistemology.

Epistemology asks how we can know reality given only our subjectively experienced perceptions and disciplined investigations of correlations among those perceptions. Those disciplined investigations make predictions (hypotheses) about perceptions and test those predictions by controlling those perceptions; they treat some CVs as axiomatic postulates or as givens resulting from epistemologically prior sciences (as mineralogy depends upon physics and chemistry, or pharmacology depends upon chemistry and physiology).

A science is a system concept and its CVs include its postulates, methods, data, conjectures, hypotheses, theoretical entities, models, etc. in addition to the immediately observed objects and relations of the science. Control of a system concept entails that control of any of its inputs does not disturb control of any other of them. This is called internal consistency. Such reciprocal conflicts are seen in control of language at a much lower level, where control at nearby points in the pronunciation of a word are disturbances to one another (coarticulation effects).

Every well-established control loop in a hierarchical control system makes predictions about its CVs which it tests every time it controls its perceptual input. Failure of an accustomed means of control recruits remembered alternatives and analogous alternatives. (Analogy is a function of associative memory.) Continued failure to control an important CV leads to reorganization until error in the hierarchy is reduced. Because of reciprocal disturbances (‘co-control effects’?) reorganization in one place can have wider effects. Reduction of error can also be accomplished by avoiding perceptual input that cannot be (well) controlled, or perceptual input that is the occasion of such conflicts when we control it. The process of acceptance of new developments in science is an example that we know well.

A cosmology is a high-level system concept with inputs from a CV called ‘myself’ and other system concepts. Theories of the sciences that one knows are important perceptual inputs to one’s cosmology. In general, each science builds upon epistemologically prior sciences plus postulates, axioms, or assumptions taken from logic, mathematics, and some kind of epistemology that claims the veridicality of (disciplined) observations. In principle this gives all the sciences a common basis in physics and chemistry, as is sometimes claimed here for PCT. Unifying the sciences has in fact been challenging. A major attempt was made in the 1930s with the International Encyclopedia of Unified Science, which didn’t get beyond its first volume.

The way forward at higher levels is in the investigation and modeling of social phenomena. Those are matters that humans find most interesting and important to them. One may say that the most important variables are in the bottom layers of Maslow’s pyramid, but in our world control of food, shelter, etc. requires social competence, and this has been true for many millennia of our ancestors. Social phenomena are collectively controlled, that is what makes them social. Not all of us are interested in investigating and modeling social phenomena, and humans are unequally skilled or endowed to control perceptions of them (viz. Asperger’s syndrome and autism).

Yeah, it does seem pretty postmodern and ridiculous. Still, I can’t find a consistent story that would explain the origin of our knowledge and have an objective test of truth. Everything has to be subjective, and if there is something outside of our experience, and I assume there is, we can’t say much about it. Saying there is “something” is already too much, because it is not “some thing” - “things” are creations of minds. When we talk about reality, we are always just referring to other levels of our experience.

there is something we can know about reality and about the relationship between perception and reality. (This opinion can be justified for example starting from the idea of the evolution of living beings, like Martin has often stated.) We cannot know the exact mathematical character of that relationship but we can assume that it is – at least often – an analogical dependency functio

I don’t understand that. How is evolution related to the relationship between perception and reality? I can see that the development of science shows that we can get progressively better approximations of what is happening between two observed variables, we find a law that transforms one to the other. I don’t see it in biological evolution.

I do agree with that, certainly a valid area of research, but the perceptions are still subjective - each person has their own, and we won’t get far if we make mistaken assumptions about what is real and what is perceived. :man_shrugging:

Objectivity: The closest we get is intersubjective agreement. Science does this in a rigorously disciplined way, ideally. That varies from one field to another, and is most successful when an important part of the discipline is to work to find problems with one’s favored hypothesis.

Evolution and the veridicality of perceptions: organisms that construct and control perceptions that cannot be controlled by reliable environmental feedback paths (whatever they may be, and whether or not links in those paths are themselves perceived) are less successful and have fewer progeny that come to reproductive maturity.

It’s not a simple relationship. One aspect of control is to stay within an environment that provides reliable feedback paths, and in very many cases, certainly all social, communal, colonial, etc. organisms alter the environment to increase the number and stability of the environmental feedback paths by which individuals control their perceptual inputs. But this, too, is a product of evolution, i.e. individuals participating in this are more successful in producing offspring that come to reproductive maturity than are individual organisms of the same type fending for themselves.

PCT discussions have been uncomfortable with the evolution of more complex organisms with control systems made up of simpler organisms, and the evolution of pathogens and parasites to symbiotes and then to integral components of those complex organisms. The facts are incontrovertible; the obvious application to our human condition and future is not. We may kill everything off before our communities are reliably symbiotic rather than pathogenic or parasitic.

Have a look at PPC I.2.1and 1.12.1. In case you don’t have the link to PPC Vol I, here it is https://www.dropbox.com/s/aj9g3njbcgqkowc/PPC_Vol_1.rtf?dl=0

I think my reply six months ago was rather dismissive and not conducive to development of a common understanding. For that I apologize.

My perception of the core issue is to what extent perception could possibly model real reality. In the abstract, this seems like an impossible question to answer, but I thought that my reference to evolution made the difference between modelling the functionality of real reality and modelling the mechanism, the “things” that produce the functionality, explicit. We can make (in the perceptual side of the perceptual control hierarchy) structures that I have called (following Wiener) “White Boxes” that do to any level of precision what we observe real reality to do, but we no more know how real reality does it than we can know what language an “object” in object oriented coding is written if we are given its complete specifications for what its output should be given any input.

Just as the object-oriented coder can choose any coding language in which to implement the specified object, so real reality may use any method (including my network of bureaucratic gnomes) to do what our observations and experiments show it to do. We can know the specifications, but not the code.

Evolution comes into this picture because of the assumption that the methods used by a child are almost always the same as those used by the parent(s), if the child survives long enough to become a parent. Simple organisms have many children, most of which will not survive to have children of their own. The more complex the organism, the fewer children, and the more the parent influences the child’s environment to enhance its likelihood of surviving to reproduce.

We my not know how real reality does it, but we can learn what it does to any level of approximation. Physics goes a long way in chasing ever finer approximations to what real reality does, as physicists invent newer variants of white boxes within white boxes. That, too, is what evolution does in creating an ever increasing number of interacting species of actors, from fungi to trees, bacteria to fish and mammals.

Evolution functions through its effects in real reality, but we can imagine white boxes in white boxes to model its functions. Never can we discover how the corresponding black boxes do what they do.