What would be the point of all this modelling without its ability to promote accurate, relevant action - pulling causal levers whose structures we partially represented in the world? — fdrake

Assumption that there is ‘a point’. Pointless question as far as I can tell. What does i mean to ‘have a point’ or ‘be pointless’? We can ask what the ‘theoretic’ attitude is doing, how it is structured and how theories change.

You could just settle on the hidden answer to your question and say ‘accuracy’ is the point - that was essentially what you implied. I ask what else could a ‘model’ do other than refine itself in order to increase accuracy? If increasing ‘accuracy’ is the be all and end all then some ‘error’ is an optimal means of exploring beyond the immediate bounds ‘known’. We don’t purposefully make mistakes, yet we certainly can, and do, learn from mistakes brought about by ‘inaccuracies’.

All of this needn’t be led under the assumption of some abstract ‘input’/‘output’ system. That is merely an expression of the ‘theoretic’ attitude - a necessary means for distinction among the ‘white noise’. For further more ‘substantial’ evidence for this neural priming is an expression of this in the ‘natural sciences’.

If you have a goal of opening a door, your hand position needs to adapt to where the handle is and how it works; there's an accuracy constraint involved with the door's location, functionality and so on; irrelevant of how it's split up into perceptual features. — fdrake

Yes, absolutely, so we definitely have to have a real world otherwise our models are modelling nothing, our entropy resisting organisation has to have some entropy to resist - we agree so far.

Focus for a moment on the γ, if I've read things right these are "hidden causal states" which are later associated with environmental parameters θ rather than available sufficient statistics x, again if I've read it right. They are hidden, but inferred upon by the whole active-modelling process under some representation ("recognition dynamics"?). — fdrake

Yes, that's my reading of it too - which is good because I'm trusting you to understand the maths better than I do.

Our ability to act well in an environment depends upon having a good model of it; to update our model through some error minimising in response to our current goals and current environment; the model doesn't just take input from previous modelling steps, it takes input from external states with their own dynamics under some processual representation. It would be unable to guide action if it did not have a satisfactory (sufficiently accurate) representation* of the external states as they are relevant to our goals and bodily constraints. — fdrake

Yes, still in agreement here.

I think you're emphasising that the starred representation* is another flavour of model; which it is; but it's also observation process of relevant structures for us in the environment. What would be the point of all this modelling without its ability to promote accurate, relevant action - pulling causal levers whose structures we partially represented in the world? — fdrake

OK so this is where we part, but it's quite nuanced, so I'm going to be as specific as possible.

1. I don't see anything in the maths (and this, I think is what you've been looking at, but I'm not convinced) that requires there to be distinguishable 'structures' in the environment. Only for there to be heterogeneity in the environment (otherwise it would be impossible to model, plus with no randomness, there'd be no probability gradient to work against). All I'm arguing here is that all 'structures' are models of some sort.

2. You're right about 'pulling causal levers', but the upshot of the free-energy principle and it's application to systems is that there is no higher goal. Even evolution is just acting on us as a higher order free-energy minimising system (Friston talks about evolutionary free-energy minimising in one of his lectures, but I can't find any writing on it I'm afraid). So we don't have a primary goal of survival in the environment, say. We have a primary goal of surprise reduction, as does evolution (as a system) and that imposes the goal of survival on us. so...

3. It leads us back to the comment that Bayes optimal solutions have no need for accuracy in the environment, only for variance minimising within the system.

I've used this example before, possibly, but Gravity is a really good one. Gravity (it seems) is totally wrong, completely not how things actually might be according to some very coherent alternative models. But it's a really effective model for how to get about on Earth. Completely inaccurate - very effective. It's quite possible, if not probable, to have completely inaccurate models which serve us very well in terms of variance minimising, and they will remain in place, even in favour of more accurate models (which might be too energy-hungry to actually use)

Assumption that there is ‘a point’. Pointless question as far as I can tell. What does i mean to ‘have a point’ or ‘be pointless’? We can ask what the ‘theoretic’ attitude is doing, how it is structured and how theories change. — I like sushi

OK, rephrase it:

"How do we account for the presence of a modelling process that samples our environment and our actions based on causal stuff that's relevant to us (part of our model) in a manner that maximises our accuracy of representation of modelled content without requiring that the modelled content itself is informative of the structure of an external (unmodelled!) world?"

1. I don't see anything in the maths (and this, I think is what you've been looking at, but I'm not convinced) that requires there to be distinguishable 'structures' in the environment. Only for there to be heterogeneity in the environment (otherwise it would be impossible to model, plus with no randomness, there'd be no probability gradient to work against). All I'm arguing here is that all 'structures' are models of some sort. — Isaac

Mmmm... I don't think there's any specified structure in the paper. But there's ghosts of it, suggestions. The external dynamics model $f$ and the sensation model $g$ are "continuous nonlinear functions of (hidden and causal) states, parameterised by $\theta_i$" (where $i$ is the hierarchy level), they have a specific functional form. So there's some structure there, it's not "some arbitrary function", it's "some definite capacity of a human relating to a world" (albeit a very involved one). I'm thinking of $f$ and $g$ as "world->self" interfaces in terms of perceptual/sensation content and "self->world" interfaces in terms of actions taken.

So when you say:

Only for there to be heterogeneity in the environment (otherwise it would be impossible to model, plus with no randomness, there'd be no probability gradient to work against). All I'm arguing here is that all 'structures' are models of some sort. — Isaac

This gets mopped up by the error terms, no? There's a signal/noise distinction operative; the signal is encoded in the $f$, $g$ (in terms of external state "recognition dynamics" and sensations), the errors from these are current environmental or bodily heterogeneity.

I think you're interpreting the only site of interface between the agent and the "unmodelled" external world as the error terms (unstructured heterogeneity)? Whereas we might be able to representatively sample its dynamics using only our $f$ and $g$ models. I'm kinda reading this whole process as a machine for representatively sampling actions/sensations/perceptions from a dynamical causal model of relevant causal factors using available information (sufficient statistics).

Again, if I've read it right, $f$ and $g$ are not time varying functional forms: they embed one pattern which is evaluated at different arguments (like evaluating f(x) = x at x=2 and x=3); so they're individual level relational styles of individual to world. They're "the way our bodies weigh the causal structure of our bodies and our environment". I'm inclined to believe (1) that there's room for discovering real structure in the external world so long as $f$ and $g$ are representative (within our narrow window of concern/narrow scope of perceptual features) (2) that it makes sense to consider $f$, $g$ as embedded in an evolutionary process that would be surprising if "an organism's ability to evaluate relevant causal structure to it accurately" was not selected for (in some circumstances) - going back to my camouflage example from earlier.

Prosaically: sometimes the relationship between two quantities (and their represented phenomena) really is indistinguishable from a straight line. And this is exploitable.

Well, it's because certain properties of experience are perceiver-dependent and not in the objects themselves. The air feels cold, but that doesn't mean it is cold in an objective sense (the feeling of cold varies between individuals and time). It just feels cold to you now. — Marchesk

So subjective experiences are simply beyond objective understanding. Something like non-overlapping magesteria that Stephen Jay Gould proposed to keep religion and science from each other's territories.

I guess we could proceed along those lines and say consciousness is literally incomprehensible because just like we can't separate the subjective from consciousness we can't take away objectivity from understanding/comprehension. After all comprehension has to be unbiased which is another name for objectivity.

Do you agree then that as per Nagel's argument, consciousness is, by its nature, incomprehensible? Or is there the better option of a type of comprehension that is non-objective which can access knowledge that objectivity can't handle?

All of this needn’t be led under the assumption of some abstract ‘input’/‘output’ system. That is merely an expression of the ‘theoretic’ attitude - a necessary means for distinction among the ‘white noise’. For further more ‘substantial’ evidence for this neural priming is an expression of this in the ‘natural sciences’. — I like sushi

This would probably derail us, but I can't help but indulge.

As if any approach to conceptualising anything was not a 'theoretic attitude'. Assuming you're coming at this from a phenomenological angle, the epoche only becomes relevant within an interpretive framework; its own flavour theoretical attitude. Spiralling out into its own mode of inquiry. No reason for primacy here.

I’ll stop banging my head against this wall I think

I guess I should simply start my own thread and see if I get any traction there instead.

Thanks :)

This gets mopped up by the error terms, no? — fdrake

I don't think so, because the error terms are related to the primary sensory input (box2 in Friston's paper) which are themselves non-updatable models (the model here being determined by the architecture - signal x being built from input y), the function only takes x (and errors related to priors as to the causes of x), not y.

we might be able to representatively sample its dynamics using only our f and g models. — fdrake

We might, agreed. But could we then feedback errors? The sensory inputs are one way, the first internal feedback loops are in the perception cortices. We can correct errors to say "that's not a grey square, it's just a shadow", but we can't update the retina with "that number of photons didn't hit you", or whatever 'real' feature we're tracking. So it comes back to the mechanism by which this update, reduction of errors, could take place. And without updates, we're unable make any claims to variance reduction towards any 'structures' which might be there.

We might, agreed. But could we then feedback errors? — Isaac

I think so?

I'm imagining $f$ as causal indicators under some model. Let's say that we have, like in one of Friston's talks (the one you linked me?), a task where we invite subjects (in his case simulated) to classify images of faces based on their orientation; are they upside down or right-side up? We can imagine the environmental stimulus $\gamma$ here as the light reflecting off the images or the light emitted from a computer screen, consistent with/strongly suggestive of face patterns of course. The $\alpha$ actions would be the saccades, fixations and other eye movements employed in visual search for features that indicate the orientation of the face. But what's the $f$?

Human subjects clearly have a sufficiently broad and tailored $f$ that allows us to do this task quickly and reliably. If you record eye movements of human subjects looking at faces, we look at eyes, nose, mouth preferentially. I think this has something to do with $f$ as it relates to this task.

A feature extraction/detection algorithm was used to make simulated agents learn a good $f$ for this task; IE, they learned what parts of the face to look at to quickly determine its orientation. IIRC the simulated agents were eye movement patterns, and the $f$ they learned were the representative features of the face for the orientation determining task. These turned out to be noses, eyes, mouths. You don't need to sample (eye fixation points + saccades) much from this feature space (eye related pixels, nose related pixels, mouth related pixels, or their associated characterisation in light frequency and apparent source) to determine the orientation of the face. These points of high information density relative to the task become causal indicators for the face orientation. If (eyes + nose + mouth look like this) then (face should be that).

You can always make an error in perception, or in a sub-perceptual task like allocating a part of a face to a facial perceptual feature. In terms of "what's not in the trend (the error) assuming $f$, the stimulus $\gamma$ and the action $\alpha$?", we might be in a state of momentarily misperceiving that the face has one eye due to only sampling from one half of it, say, (these things seem to execute prior to there being a phenomenal character associated with the face), or we might judge a shadow on one side of the nose as a discolouration of it, and this error promotes another saccade $\alpha\prime$ to check for the other eye or other bits of the nose. We often saccade from eye to eye when first viewing a face (used to work with eye movement data, dunno how heavily weighted eye-eye saccades really are in facial recognition tasks). This is just conjecture, of course!

At any given "step" in these simulated agents' classification algorithm, they'll have an error associated with the probability of the face being upside down or not (a conditional expectation of the face being upside down given what's been seen, the internal state, the sensations and the actions). You can minimise energy expenditure and maximise accuracy by focussing on information rich features; if the algorithms, and we, tried to guess the orientation of a face solely by sampling from bits of bare skin on the nasal septum, we'd do a much worse job probably, as a small section of bare skin in the middle of the face that's not part of a facial feature is probably sufficiently reflection invariant that it won't inform either way.

So for the classification task, we can split the face up into the causal features of nose, eyes, mouth, and minimise the errors from there. Edit2: $f$ is probably really really broad but gets channelled into distinct behaviours by the history of the system. I don't think $f$ is updating for us, it's more that how we evaluate it changes over time, relative to what we've learned, what we're doing, and what our environment is (all the history embedded in our current action).

Edit: if eyes + nose + mouth seem arbitrary as causal indicators for facial orientation, their relative positions encode a lot. The nose is a roughly central point irrespective of the orientation of the image, and if you look up on the image and see a mouth, the face is upside down, if you look up on the image and see an eye, the face is right side up. And vice versa. The whole orientation of a human face is determined by the relative positioning of the eyes and nose and mouth; so it is not surprising that these features are learned for the task. If you wanted to represent this as (pixel colours) $\gamma$ with a feature model $f$ explored with eye movements $\alpha$ and current perceptual ambiguity $w$, I think that works too?

Edit3: pretty sure I'm confusing salience and information density here, though the two are related (I'm guessing, there's some reference to the distinction in the Friston paper, only worked with something related to saliency maps before).

Did you not read what I wrote, or not understand it, or not agree with it. If you're not going to actually respond in any way to what I write there's little point in continuing is there? — Isaac

I re-read the exchange. I understood it. I didn't so much disagree with you. Rather, I found that it was rather incomplete, in that you offered choices for me to agree with, but not one that was close enough to what I hold. You're the one asking me for clarification... I gave it.

We have before us now, a listing of categorical errors...Which ones will be used to render judgment upon whether or not a premiss of our choosing qualifies as being guilty of categorical error as compared/contrasted to other kinds of errors? — creativesoul

Categorical errors can only be demonstrated by showing the falsity of the proposition from which they were originally given. I suppose one could list them, but recognizing them would seem to be sufficient.

As already mentioned, c.e. is a mistake in logical form, wherein a subject of a logical statement and its predicate do not relate, or, the subject isn’t given a predicate to which it can be related. In the case of human thought, the subject is always a phenomenal object, and the predicate is always at least one of twelve possible pure conceptions. For instance, when I perceive an object, a category of “quantity” must relate to it in the logical form “all x are y”, I must think of that x as having y belonging to it. Because I already know all dogs are canines, if I perceive an object, and I judge the concepts fur, missing frontal lobe, dew claw, chewing teeth as predicates of it, but do not judge it to be predicated by canine, I commit a categorical error by cognizing the object as a dog. It may very well have been a wolverine.

But that’s not very hard to grasp. Where the system becomes relevant is when the object perceived is unknown to me. I must follow the logical forms in order to call the resulting cognition of it, experience and thereby, knowledge. If every rose of my experience is red, upon perceiving an object with some measure of the predicates of the red rose of my experience (flower, scent, petal shape, thorns, etc.) but is not red, I am justified in denying it is a rose at all. The dog example is a judgement of category “quantity” using the subset “universal” in the logical form “all x are y”. When it is proven to me the object I perceived is a rose, but not a red rose, then the category “quantity” invokes the subset “singular” in the logical form “this x is y”, and from that, I am justified in thinking the third subset of “quantity” with the “particular”, in the logical form “some x are y”. Henceforth, I should have no logical inconsistencies in cognizing roses of different color.

In 1904, trains, tracks, terminals, lightning, tape measures, pencil, paper, and a myriad of other real objects were already well-known. When ol’ Uncle Albert was whiling away the hours at his ramshackle desk in his ramshackle office, he put two and two together and came up with a four nobody had ever noticed. It would have been absolutely impossible for him to think any of what eventually came to be SR without the “existence” of those listed objects, without the “possibility” of the objective validity of his thinking, without the “relation” of cause and effect, and without the “reality” of the end result, which in this case was merely mathematical. The “existence” of SR’s objective reality, over and above its objective validity, had to wait for technology to play catch-up, but that does nothing to diminish the fact the whole thing was the product of sheer imagination, that is to say, predicated on empirical means but having no empirical ends whatsoever.

The categories of understanding (SR) are the logical conditions for pure thought, or, a priori cognition, and thus possible experience. The categories of judgement (roses) are the logical conditions for empirical cognition, or, experience. Both are themselves pure a priori rational faculties, therefore their respective products are far and away antecedent to linguistic conditionals. It follows that categorical error is simply the non-employment of the proper category with respect to an object, such that the thought of it is contradictory, or the cognition of it is false.

As an aside, these categories are my rendition of those things which we had previously agreed must exist in their entirety before we are permitted to make our correlations in what you call thought/belief.

, you haven’t really learned what you set out to discover, insofar as you’ve got the “how” but not necessarily the “how come”, because your subject himself may not even know.
— Mww

I don't think the 'how come' is a measurable thing by any metric, so it's not a relevant investigation. Pick your reason. — Isaac

I submit there is at least one metric for measuring at least one “how come”, and that measurable metric is behavior, with respect to the “how come” called morality. The former is directly proportional to, and directly dependent on, the latter. While behavior may only be observational, and hardly scientific, it can still be measurable by relation to some other behavior. And it would certainly seem to be relevant, considering the general proclivity for humans to piss each other off.

So, we must surely abandon Kantian language here. For anything that exists in it's entirety prior to our naming and descriptive practices qualifies for that which exists in and of itself(Noumena). — creativesoul

The categories are not phenomena, there is no object that can be thought for them. That which is not phenomena is not thereby automatically noumena. Noumena, if they be given at all, are only given from an intuition different than ours. Dolphins may think with noumena rather than phenomena....we don’t know, and unless we communicate with mutual intelligibility, we won’t.
——————-

Furthermore, we must have knowledge of both our premisses and that which exists in it's entirety prior to our premisses in order to perform a comparative assessment between the two. That comparison is required in order to know that one has indeed committed the error called "categorical". — creativesoul

Premises in syllogisms or propositional logic in language form, yes, but the categories enter the cognitive stream way before knowledge, which makes explicit categories are not knowledge-apt. Categories are conditions, and we know them only as conditions, only as part of non-linguistic logical relations alone.

We have before us now, a listing of categorical errors...Which ones will be used to render judgment upon whether or not a premiss of our choosing qualifies as being guilty of categorical error as compared/contrasted to other kinds of errors?
— creativesoul

Categorical errors can only be demonstrated by showing the falsity of the proposition from which they were originally given. I suppose one could list them, but recognizing them would seem to be sufficient. — Mww

There is more than one acceptable sensible conventional sense of the term "categorical error". It is a name with more than one referent. The one will we choose as a standard to render subsequent judgment concerning a candidate of our choosing will directly determine, establish, and influence the judgment call. The sense sets the parameters. Not all senses of "categorical error" are commensurate.

So...

Why are we talking about categorical errors, when what counts as a category is completely and utterly determined by us?

I seek to discover that which exists in it's entirety prior to our discovery process, therefore prior to the naming and descriptive practices commonly called common language use...

-----------------------------------------------------------------------------------------------------------------

Noumena is untenable. I've already offered adequate argument for that conclusion. It's been left sorely neglected.

I'm not actually sure we mightn't just completely agree, so far as your actual line of thought on this extends, but I feel like I'm missing some connection, I may be just slow to join the dots, and if so I'd appreciate some clarification, but apologise if it just goes over old ground. To that effect, I want to re-phrase what I think we're trying to speculate on (just in case that's got lost along the way, and that's why I feel like I'm missing something). We're looking at whether, and how, actual structure (real existent patterns in the hidden states) from outside the Markov blanket of an organism are reflected in the models of those hidden states caused by the variance minimisation of the Bayesian functions inside the Markov blanket? If they are, then we can say there's likely to be structure in the hidden states, and, crucially, by structure I mean that they are some way, and not any other (as opposed to simply that they could be 'seen' as some way, but also could be 'seen' as another - which would be model dependant realism).

I know it's not the most complete of replies to your thorough exposition, but I feel like I might be going off on the wrong tangent entirely if I'm wrong about the above.

The reason I'm asking is because you've given pretty much the same understanding as I have in my mind of the process, yet I've drawn one conclusion from it, and you another. The salience modelling you describe is, as we've discussed, definitely a strong indicator that the underlying hidden states outside the Markov blanket must be heterogeneous (here in terms of light reflected), and I think you're right that information density and salience are linked (although see the discussion of edge detection later in the paper, for the difference. Edit - it's not in that paper is it, I've just looked, must be another one). But still that's all we're getting, heterogeneity, not fixed structure (the hidden states being this way not that way).

Let's say our hidden state is h (I can't do that fancy mathjax stuff - you'll have to guide me to some instructions sometime), and we give it some structure represented by a fixed value set, so h={6,10,19,108,4,9}, and this set fully describes the structure of h. Our f is modelling h by sampling. It's doing so efficiently, based on prior densities, updated by inputs from latest samples benefiting from some error recognition. If this were happening purely mathematically, we should fairly quickly end up with the sample set. But...

Firstly - If our priors for the input mechanisms are not expecting numbers above 100, there's nothing coming in to the higher level models (working on what caused the lower level ones) to feedback to get the input mechanism to update its priors to allow for the possible recognition of numbers over 100. We cannot get at the primary input to update it because they're hardware, not software. Their priors are the equivalent of drivers, or firmware, and so don't have the flexibility to be updated to account for any and all potential structures of hidden states outside of those they were created to detect, and, to meet your point about evolution, the physical constraints of the stuff they're made from.

(complete aside here, but this touches on the work Iain Stewart did many years back about morphological constraints vs. natural selection as being far more determinant of final organisms - but that's only tangentially linked)

Secondly - If we were to 'hone in' on h being {6,10,19,4,9}, that's not a wrong pattern, it's an incomplete one. Any interface with h won't be in error presuming that pattern, it simply won't exhaust the possible patterns. Like pointing to the constellation Orion and describing the pattern of the hunter with his belt and bow. It's not an error because the pattern is there, it just doesn't exhaust the possible number of patterns which no less error value.

I understood it. I didn't so much disagree with you. Rather, I found that it was rather incomplete, in that you offered choices for me to agree with, but not one that was close enough to what I hold. You're the one asking me for clarification... I gave it. — creativesoul

I understand that, but I gave you what seemed to me to be the only options (perhaps I should have made that more clear), so any response which simply re-iterates your original position without explaining how you circumvent the issues I raised seems to contribute nothing to our mutual understanding of the issue.

You might have to clarify how you're distinguishing "how?" from "how come?" here.

I understand that, but I gave you what seemed to me to be the only options (perhaps I should have made that more clear), so any response which simply re-iterates your original position without explaining how you circumvent the issues I raised seems to contribute nothing to our mutual understanding of the issue. — Isaac

As if I'm obligated to answer for issues you've raised that have nothing to do with my position...

Happy modeling!

:smile:

------------------------------------------------------------------------------------------------------------------

As already mentioned, c.e. is a mistake in logical form.. — Mww

Logical forms are existentially dependent upon common language use. Common language use... rudimentary thought and belief. Logical forms... rudimentary thought and belief...

If a square can only fit through a square hole and it passes over a circular hole, it may as well have passed over no hole at all. Or you could say where there is no square hole there is no hole. Further why say ‘square’ as ‘hole’ would suffice. Otherwise we could end up saying there are circular holes everywhere that we cannot comprehend which for the square is no different to saying there are only square holes.

Even our speculative capacity is limited. Language does have a habit of throwing up sentences that possess little to meaning.

If you don't mind I'd like to request a clarification.

What exactly does one mean by subjective experience.

I read Nagel's paper and he doesn't define subjective experience (SE) anywhere in paper. If I understood correcty Nagel's "single point of view"/SE is critical to his argument but what exactly does SE mean?

It can't be self-awareness because that's something every consciousness has so, according to Nagel's rules, can be objectively evaluated.

Is SE dependent on the milieu of our minds - mental objects like concepts and its derivatives and our physical surroundings insofar as they affect SE? If yes then we can be objective about them. We can, after all, be objective about concepts and whatever follows from them. Also the physical world too is amenable to objectivity.


If all what I said is correct then SE is an instance of our consciousness interacting with its environment, both of which it's possible to be objective about. Ergo, we can be objective about subjective experience.

Not all senses of "categorical error" are commensurate. — creativesoul

Correct, which is why I mentioned Gilbert Ryle. I figure error in semantics or error in reason are the only two worth talking about. And because language philosophy is (insert pejorative terminology here), the only error worth a damn in philosophical discourse is grounded in pure reason.
——————

what counts as a category is completely and utterly determined by us — creativesoul

I suppose that’s right enough, although I would offer that we determine whatthey are, but not that they are; the former presupposes the latter. Nonetheless, all philosophy is theoretical, and if it should be the case that a foundational tenet of a particular theory is given as merely a condition for that which follows from it necessarily, and disqualifies such condition from any empirical determinant for it, it must be considered as existing in its entirety prior to being named as such, by the rational agencies that employ it in the normal course of its mental events, in accordance with the theory. The onus then falls on the opponent of the theory, to falsify the tenet by arguing successfully in the negative. He does his dialectical opposition no justice by merely claiming fallacious rationality used in the construction of the theory, but rather, is required to give qualifying justifications for it. It may be interesting to note that, at least since Aristotle, no one has been able to promote sufficient reason to nullify the need for, and function of, the categories. Doesn’t make the theory or its tenets fact, or even irrefutably the case, but standing the test of two millennia of argument is still pretty damn good.
———————-

I seek to discover that which exists in it's entirety prior to our discovery process, therefore prior to the naming and descriptive practices commonly called common language use... — creativesoul

Cool. Let me know what you find? We can compare it to what has already been found. Or found acceptable, at any rate.
———————

For anything that exists in it's entirety prior to our naming and descriptive practices qualifies for that which exists in and of itself(Noumena). — creativesoul

Noumena is untenable. I've already offered adequate argument for that conclusion. It's been left sorely neglected. — creativesoul

In the first, the parenthetical suggest a “for instance”, as in, “that which exists in its entirety exists in and of itself, like for instance, noumena”.

In the second, noumena are claimed to be untenable, which of course, they are, for us.

By association, it follows that because categories exist in and of themselves, they are untenable noumena. This is false. Or, “That’s not only not right, it’s not even wrong!!” (Thanks, Herr Pauli!!!) Categories are neither untenable, nor noumena. In addition, I haven’t seen on these pages any adequate argument for noumena being untenable except from yours truly. Actually, him truly, me being the poor messenger. And because I probably would have agreed, and so remembered, if you’d presented adequate argument, I’m going to go ahead and say you haven’t. But surreptitiously with my fingers crossed.
——————

Logical forms are existentially dependent upon common language use. — creativesoul

No, I think not. Logical content is dependent on language use; logical form, re: the Greek laws of comprehensible, rational thought, are not:

“....Pure intuition consequently contains merely the form under which something is intuited, and pure conception only the form of thought in general. Only pure intuitions and pure conceptions are possible a priori...”

A = A, e.g., is the pure form the content must take when language fills out the form, to accomplish comprehensible, rational thought, and thereafter, comprehensible, rational communication. You can fill out the job application with any ol’ information you want, but your paycheck might end up in Alaska, just as you can fill in the logical forms any way you want, but your cognition “bird” might be everybody else’s cognition “bathtub”.

We can move on, if you like.

What exactly does one mean by subjective experience. — TheMadFool

There are several things to the definition. One is any experience which varies between individuals. The room feels hot to you, cold to me, and fine for a third person. The experience of temperature is subjective. If we wanted to measure the room's temperature, we use a thermometer which gives us an objective value which does not vary.

Another is private. I have a dream, and although I can tell you about my dream, you cannot experience the dream yourself. The experience is private to me. So although dreams can be studied objectively, the experience itself is only available to the individual who has that dream.

A third is perceiver-dependence. This is based on the kind of perceiver, and their sensory capabilities. So humans experience the world through five senses of an upright walking ape, with differences among individuals due to color blindness, being able to taste a certain chemical, incapacity, etc.

The perceiver-dependent qualities of human subjective expereince would be those sensations we have good reason to believe are generated by our nervous system, instead of being properties of the world around us. So shape, size and location are objective properties of things in the world, while color, sound, taste are properties we experience because of the kind of creatures we are. Going back to the room temperature, our experience of heat or cold is a perceiver-dependant quality. The temperature is objectively the kinetic motion of particles moving about, and not a feeling of coldness or heat.

Nagel makes the argument that science creates a view from nowhere that has no perceiver-dependent, private, perceptually-relative sensations. There is nothing it's like to be a wavefunction or a supernova or evolution. It doesn't feel like anything, it doesn't look like any color, it doesn't sound like anything. The particles moving about in a room don't feel cold or hot. Ultimately, it's mathematized models of some reality divorced from our experience of it.

No, I don't think I would, but I get what you're saying. I don't think proximity to reality measures the usefulness of the model. As such, I think it's theoretically possible that a model might be useful without relating to anything at all, but I haven't thought about that much, so my intuition may well be wrong. Interesting question. — Isaac

I’m sorry if I’m taking your point out of context but I target it because it reminds me of the Hoffman argument, outlined in his TED talk. I don’t understand how the model can completely be unrelated to reality at all. Of course it is made for organism-relevant features, but surely you’d agree wavelengths picked up by the retina are coming from reality, and surely, at least the structural regularities in experience, are ‘real’, even if they’re not fundamental or reference frame independent. There’s a broader question of whether we are epistemically restricted to the point that we can’t intuit fundamental features of reality. I am still unsure if we can but I think there’s is evidence maybe we can. The fact, for example there is experimentally robust evidence for indeterminacy at smallest scales. indeterminacy implies no underlying mechanism or principle that results in the phenomenon. It would be hard to imagine or define what something more fundamental could be once you’ve reached a level where there are no deterministic principles. So maybe we have enough access to identify fundamental features or principles of the parts of reality we can interact with. What are your thoughts?

Two questions...

What are logical forms taking account of?

Would you agree that "that which exists in it's entirety prior to common language" is a category?

the only (categorical)error worth a damn in philosophical discourse is grounded in pure reason. — Mww

Pure reason? As in reasoning from an armchair?

There are several things to the definition. One is any experience which varies between individuals. The room feels hot to you, cold to me, and fine for a third person. The experience of temperature is subjective. If we wanted to measure the room's temperature, we use a thermometer which gives us an objective value which does not vary.

Another is private. I have a dream, and although I can tell you about my dream, you cannot experience the dream yourself. The experience is private to me. So although dreams can be studied objectively, the experience itself is only available to the individual who has that dream.

A third is perceiver-dependence. This is based on the kind of perceiver, and their sensory capabilities. So humans experience the world through five senses of an upright walking ape, with differences among individuals due to color blindness, being able to taste a certain chemical, incapacity, etc.

The perceiver-dependent qualities of human subjective expereince would be those sensations we have good reason to believe are generated by our nervous system, instead of being properties of the world around us. So shape, size and location are objective properties of things in the world, while color, sound, taste are properties we experience because of the kind of creatures we are. Going back to the room temperature, our experience of heat or cold is a perceiver-dependant quality. The temperature is objectively the kinetic motion of particles moving about, and not a feeling of coldness or heat.

Nagel makes the argument that science creates a view from nowhere that has no perceiver-dependent, private, perceptually-relative sensations. There is nothing it's like to be a wavefunction or a supernova or evolution. It doesn't feel like anything, it doesn't look like any color, it doesn't sound like anything. The particles moving about in a room don't feel cold or hot. Ultimately, it's mathematized models of some reality divorced from our experience of it. — Marchesk

I agree that subjective experiences are unique and no two are alike. What I'm suggesting though is that subjective experiences are constructed out of material that can be objectively analyzed.

1. Consciousness is universal in form. (There's no difference in the nature of consciousness between two conscious beings)
2. Subjective experience is necessarily unique and can't be something that is shared
Ergo,
3. Subjective experience is not consciousness
Ergo.
4. Subjective experience is the interaction of consciousness with its environment
5. The environment of consciousness consists of ideas and the physical world insofar as it affects consciousness
6. Ideas and the physical world can be objectively analyzed
7. Consciousness can b e objectively analyzed
Ergo.
8. The interaction between consciousness and its environment can be analyzed objectively
Ergo,
9. Subjective experiences can be objectively analyzed

If one wants to know what it's like to experience X, one must know what all experience has in common. It is only knowledge of that sort that allows one to offer a subsequent sensible answer to each and every example thereof.

Experience is subjective in that it is perceiver dependent and influenced by individual particular circumstances. Experience is objective in that it consists - in very large part - of the experiencing creature's own thought and belief about what's happening at the moment, and part of what's happening exists in it's entirety prior to becoming a part of an individual's experience.

So...

The objective/subjective distinction is rendered inherently inadequate in that it's use cannot take proper account of what all experience consists of.

Discard it.

...our experience of heat or cold is a perceiver-dependant quality... — Marchesk

Experience is a quality?

Consisting entirely of Quale?

Experience is a quality?

Consisting entirely of Quale? — creativesoul

I don't know, but it's something perceivers generate in the act of perception, memory, imagination, dreams, hallucinations, etc.

Experience is a quality?

Consisting entirely of Quale?
— creativesoul

I don't know, but it's something perceivers generate in the act of perception, memory, imagination, dreams, hallucinations, etc. — Marchesk

I would agree if we changed that slightly to "help generate"...

What's a "perceiver"?

I would agree if we changed that slightly to "help generate"...

What's a "perceiver"? — creativesoul

Living organisms with active nervous systems and sensory organs.