Logic vs emotion

[drkitten]

Sorry, I hadn't realized this discussion had re-opened.

As I recall, neurons, in general, abide by the "all-or-none" law. Either they fire or they don't. However, "frequency of axonic transmission" is a variable.


I contend that this design is sufficient for a basis of all brain processes - including "logic and emotion."

Well, the first statement, in general, is a tautology. It's certainly the case that a neuron is either firing or it isn't, and it's similarly true that a car is either running or it isn't, that a bag is either empty or it isn't, et cetera. The key question is whether or not two states that can be broadly categorized as "firing" can be further subdistinguished.

And, of course, they can, as you yourself point out; two neurons can both be "firing" but of different pulse frequencies. The Mc-P neural model explicitly ignores this factor and treats alll "firing" neurons as mathematically and computationally identical. If you assume that neurons can fire at different rates and that these different rates can be distinguished, you've more or less re-invented the McClelland/Rumelhart PDP neuron. But the McClelland/Rumelhart neuron, in turn, assumes that the only subcategorization that exists is frequency.

This is an oversimplification. In general, neural pulses are all of the same amplitude, but they can vary not only in frequency, but also in phase and pulse shape. Ask any signal engineer about the amount of information that can be carried in the "phase" of a signal, and you'll see how crucial this oversimplification may be.

So the problem is that the "activation" of a real neuron cannot be described only as a single real-valued number (such as the firing rate); using a real-valued activation function is a computational simplification. I could get a more accurate model by assuming a complex-valued (frequency and phase) activation function, but this would be computationally more complex, and still leave out all the information carried in pulse shape. If I were much smarter than I am, I might be able to come up with a numeric encoding of pulse shape (perhaps in terms of Fourier coefficients), but the resulting neural model would be a nightmare to evaluate. Even when I do this, I'm focusing only on the electrical aspects of signal transmission and not on the chemical.... you see how the creation of a perfect model quickly turns into a hole with no bottom.

However, only by using this more detailed model would we be able to construct artificial neural models that are able to mimic the full behavior of real neurons.

But what does this have to do with logic vs. emotion? If a human modeller develops a mathematical abstraction, and then shows that this abstraction is in theory capable of performing a particular behavior, this says little or nothing about how such behavior is actually realized in the human brain. I can, in theory, build a computer running Windows XP out of water pipes and valves, or out of the abstract "cells" that form Roger Conway's "Game of Life." The real computer on someone's desk, however, is made out of doped silicon. The fundamental reasoning is flawed :

1) The human brain does X
2) A model based on Y does X
3) A model based on Y does Z
Therefore,
4) The human brain is based on Y
5) The human brain does Z

Neither conclusion can be supported within this argument. Frankly, I don't think it matters if your framework "is sufficient for a basis of all brain processes." What matters is if your framework is an accurate description of the basis that the brain uses, something that only the neurologists can tell us.

 

[BillHoyt]

Frankly, I don't think it matters if your framework "is sufficient for a basis of all brain processes." What matters is if your framework is an accurate description of the basis that the brain uses, something that only the neurologists can tell us.

It most assuredly does matter if you can demonstrate a framework that is a sufficient basis. The demonstration would put another nail in the coffin of the various guises of vitalism. Likewise, the model can inform research directions for neuroscience.

 

[Dymanic]

Originally posted by new drkitten


This is an oversimplification. In general, neural pulses are all of the same amplitude, but they can vary not only in frequency, but also in phase and pulse shape

My previous understanding (which you have challenged...thank you) was that those have not been found to be significant factors. (Or was it that they have been found not to be? Big difference...not sure).

Anyway, I'm going to stick my neck out a little, and post a somewhat longer passage than I ordinarily would in view of the forum rules on copyrighted material. The article I linked is pretty long, and I wonder how many here will find time to slog through it all, but what Stapp is saying seems quite pertinent to these discussions. (Mods, if I'm over the line, let me know, and apologies in advance).

One section that caught my interest was the section titled: "Nerve Terminals":

Nerve terminals lie at the junctions between two nerves, and mediate the functional connection between them. The way they work is this. Each "firing" of a nerve sends an electrical signal along that fiber. When this signal reaches the nerve terminal it opens up tiny channels in the terminal membrane, through which calcium ions flow into the interior of the terminal. Within the terminal are "vesicles", which are small storage areas containing chemicals called neurotransmitters. The calcium ions migrate from their entry channels to special sites, where they trigger the release of the contents of a vesicle into a gap between the terminal and a neighboring nerve. The released chemicals influence the tendency of the neighboring nerve to fire. Thus the nerve terminals, as connecting links between nerves, are basic elements in brain dynamics.

The channels through which the calcium ions enter the nerve terminal are called "ion channels." At their narrowest points they are not much larger than the calcium ions themselves. This extreme smallness of the opening in the ion channels has profound quantum mechanical import. The consequence is essentially the same as the consequence of the squeezing of the state of the simple harmonic oscillator, or of the narrowness of the slits in the double-slit experiments. The narrowness of the channel restricts the lateral spatial dimension. Consequently, the lateral velocity is forced by the quantum uncertainty principle to become large. This causes the cloud associated with the calcium ion to fan out over an increasing area as it moves away from the tiny channel to the target region where the ion will be absorbed as a whole on some small triggering site, or will not be absorbed at all on that site.

This spreading of the ion wave packet means that the ion may or may not be absorbed on the small triggering site. Accordingly, the vesicle may or may not release its contents. Consequently, the quantum state of the vesicle becomes a mixture consisting of a state where the neurotransmitter is released and a state where the neurotransmitter is not released. This quantum splitting occurs at every one of the trillions of nerve terminals. This quantum uncertainty in what is happening at the level of the nerve terminals propagates via the essentially mechanical Process 2 first to neuronal behavior, and then to the behavior of the whole brain, so that, according to quantum theory, the state of the brain becomes a cloudlike collection of an infinitude of classically describable possible brains, each representing a tendency for some corresponding experience to occur.

What is the effect of this replacement of the single, unique, classically described brain of classical physics by a cloud-like quantum brain state composed, essentially, of myriads of alternative possible classically describable brain states?

A principal function of the brain is to receive clues from the environment, form an appropriate plan of action, and direct the activities of the brain and body specified by the selected plan of action. The exact details of the chosen plan will, for a classical model, obviously depend upon the exact values of many noisy and uncontrolled variables. In cases close to a bifurcation point the dynamical effects of noise might even tip the balance between two very different responses to the given clues: e.g., tip the balance between the 'fight' or 'flight' response to some shadowy form.

I definitely agree with what you say about mistaking the map for the territory, btw.

 

[drkitten]

My previous understanding (which you have challenged...thank you) was that those have not been found to be significant factors. (Or was it that they have been found not to be? Big difference...not sure).

"Have not been found to be significant" is more accurate; opinion is actually divided in the relevant scholastic community. However, this is also a somewhat misleading statement. Since we don't have an artificial neural model that can be used to produce anything close to human-scale performance, we really don't have enough evidence to determine what is and is not significant. In this regard, the neural modelling community is decades behind the actual neuroscientists, since the Mc-P (and even Mc-R) neurons have been shown, repeatedly, to not be capable of mimicking behavior that can be routinely observed in vivo. However, no one has yet been able to produce a behavioral "significance" to any of these problesm with the models.

So, basically, the argument runs "our neural models are good enough, so why worry?" I can see the appeal of that argument. I haven't seen any currently unsolvable problems that would be more easily solvable by making the underlying formal model computationally intractable. The interesting philosophical questions (for example, vitalism, the nature of neural "representation," or the symbolic vs. connectionist nature of language processing) are largely irrelevant to the exact details of the models. So from a philosophical point of view, either vitalism is correct and there is a fundamental reason why artificial neural networks will never work irrespective of the model details, or there is a particular framework from which qualia, perception, awareness, and consciousness will arise from an abstract collection of cooperating units -- irrespective of whether these units are Mc-P neurons or real nerve cells.

And if you can tell me which, please let me know. Unless your name is Ian, of course....

 

[JAK]

__________________________________________________

Sorry, I hadn't realized this discussion had re-opened.
...
But what does this have to do with logic vs. emotion? If a human modeller develops a mathematical abstraction, and then shows that this abstraction is in theory capable of performing a particular behavior, this says little or nothing about how such behavior is actually realized in the human brain.
...
I don't think it matters if your framework "is sufficient for a basis of all brain processes." What matters is if your framework is an accurate description of the basis that the brain uses, something that only the neurologists can tell us.

__________________________________________________

Well ... actually much of this is coming from a neuropsychologist, Dr. James Newman (http://www.bvinst.edu/faculty/~newman.htm) of the Colorado Neurological Institute (Denver). The theory is consistent with the works of Dr. Antonio Damasio, Professor and Head of the Dept. of Neurology at the U. of Iowa, as well as Paul D. MacLean, Richard M. Restak, and other neurobiologists. It is also consistent with a variety of other recent theories (Global Workspace Theory by Bernard Baars, Perceptual Control Theory by William T. Powers, Behavioral Investment Theory by Gregg Henriques, and numerous others). Brain processes mimicking natural selection (survival of the fittest) has much clinical and research support.

Even so, you are opening intriguing and exciting possibilities regarding "phase and pulse" of neural transmission. Do you have references of who is working in this area? Are you working in this area? I would be most grateful for the info.

 

[JAK]

Well ... actually much of this is coming from a neuropsychologist, Dr. James Newman ... Global workspace Theory by Bernard Baars ...

Here are a couple of quotes from Newman and Baars:

My own work has emphasized combining such evidence into a single, integrative type of organization called "contrastive analysis," which compares conscious vs. unconscious processes across numerous experimental domains. A vast amount of data along these lines is already available. Building on a contrastive analysis of the functional capabilities of conscious and unconscious processes, I have presented an integrative theory called Global Workspace (GW) Theory, a broad architectural notion of the nervous system in which conscious events are viewed as mental representations that have at least three necessary properties (a) they are global, so that their contents are disseminated throughout the nervous system; (b) they are internally consistent, because conflicting processes will compete each other out of consciousness; and (c) they are informative, in that they trigger widespread adaptation among specialized processors in the nervous system. - Baars http://psyche.cs.monash.edu.au/v1/psyche-1-06-baars.html

... neurologically this something is activation via a global attentional matrix which both: 1) controls access to consciousness by competing unconscious processors; and 2) integrates the multimodal representations of those processors which gain momentary access into unified, conscious percepts.
...
Various coalitions of unconscious processors compete for access to this global workspace (Newman & Baars, 1993).
...
(The) primary adaptive advantage is that it allows the organism to focus the processing resources of the central nervous system upon the particular stimulus most relevant to it in the moment.
...
the existence of myriad unconscious processes -- working outside of, and competing for access to consciousness
...
a single "stream of consciousness" is nature's solution to the problem of prioritizing experience, so that what is most dangerous or attractive or advantageous or interesting at that moment gains our undivided attention. -Newman http://psyche.cs.monash.edu.au/v1/psyche-1-13-newman.html

Clearly, an approach founded in competitions within the neural structures of the brain.

Nevertheless, "phase and pulse" may play a significant role in this competition.

 

[synaesthesia]

This dichotomy is very problematic and misleading.

One thinker on cogntive domain who I respect as being amongst the closest to the right scientific track is Howard Gardner. His theory of multiple intelligences (an unevolved though highly evolvable theory) integrates a working knowledge of statistics. When we divde up cogntive areas, several criterion must be considered. I believe Gardner uses eight such as

Distinctiveness of neural pathways involved; selective damagability with respect to other domains of cognition.

Plausible evolutionary history.

Distinctiveness of basic conceptual operations involved in the domain.

Now the point is not that his criterion (of which I provided only a partial sample) are immutable. Intelligences can be divided into twos (which I regard as a bad move) or into hundreds of domains, or myriad. The point is the precise heuristic value we expect of our divisions.

Howard Gardner's eight-or-so (for they are not eight in number by design or necessity) categories of intelligence are based in experimentally validated statistical models. They are very open to use in multidisciplinary thinking because we are explicitly recognizing the limitations of the heuristic.

To divide intelligence into emotional and logical (as if visual, kinethetic, intrapersonal and spatial reasoning can be easily reduced to these two, or they be reduced without the others) is a failure of imaining the sort of relationships that are held between reasoning and motivation and(?) (emotivation?).

 

[Suggestologist]

To divide intelligence into emotional and logical (as if visual, kinethetic, intrapersonal and spatial reasoning can be easily reduced to these two, or they be reduced without the others) is a failure of imaining the sort of relationships that are held between reasoning and motivation and(?) (emotivation?).

It's interesting you mention sensory modalities. Kinesthetics (external) are associated with emotion and dissociated with logic. Visuals are more detailed with logic and more fuzzy with emotion. I think they are highly correlateable.

 

[JAK]

This dichotomy is very problematic and misleading.
One thinker on cogntive domain who I respect as being amongst the closest to the right scientific track is Howard Gardner. His theory of multiple intelligences (an unevolved though highly evolvable theory) integrates a working knowledge of statistics. When we divde up cogntive areas, several criterion must be considered. I believe Gardner uses eight such as

Yes, Howard Earl Gardner’s 8 intelligences (http://www.infed.org/thinkers/gardner.htm) - linguistic intelligence, logical-mathematical intelligence, musical intelligence, bodily-kinesthetic intelligence, spatial intelligence, interpersonal intelligence, and intrapersonal intelligence.

Distinctiveness of neural pathways involved; selective damagability with respect to other domains of cognition.

Logical-mathematical intelligence and musical intelligence have been seen to be inter-related – using common areas of the brain; distinctiveness need elaboration.

I would expect damage to Wernicke’s Area and Broca’s Area to affect at least linguistic intelligence, logical-mathematical intelligence, musical intelligence, interpersonal intelligence, and intrapersonal intelligence. Again, distinctiveness needs clarification.

Plausible evolutionary history.

Plausible evolutionary history for the Baars/Newman approach can be shown from thermodynamics via Erwin Schrodinger and Stuart Kauffman.

Distinctiveness of basic conceptual operations involved in the domain.

Please explain.

... The point is the precise heuristic value we expect of our divisions.
Howard Gardner's eight-or-so (for they are not eight in number by design or necessity) categories of intelligence are based in experimentally validated statistical models. They are very open to use in multidisciplinary thinking because we are explicitly recognizing the limitations of the heuristic.

When giving creativity seminars, I, too, note differences in styles of thinking – kinesthetic, linguistic, interpersonal, spatial, etc. These can all be accommodated within Baars/Newman constructs.

To divide intelligence into emotional and logical (as if visual, kinethetic, intrapersonal and spatial reasoning can be easily reduced to these two, or they be reduced without the others) is a failure of imaining the sort of relationships that are held between reasoning and motivation and(?) (emotivation?).

Let me provide an analogy to “cellular theory.” In cell theory, bones are made up of cells, eyes are made up of cells, and blood is made up of cells. Even so, these types of cells vary in support of their tissue. Similarly, a simple neural model will allow for elegance at higher levels of structure. Instead of differing tissues, the brain has differing behaviors. Baars/Newman constructs deal with low-level brain activities which support higher ordered behaviors such as Gardner’s 8 intelligences.

Nevertheless, any of Gardner’s 8 intelligences do have success/fail attributes. Either you translate German fluently or not. Either you play a B-flat note on the violin or not. Either you add up numbers correctly and balance your checkbook or not. Regardless of the “intelligence,” there is still flow (input, process, output) as well as the success/failure of what is attempted (kick a field goal, convince a jury, perform a concert, etc.).