Being a man or a woman is understood by many to be psychological/behavioural, not genetic. If I were to somehow have my mind transplanted into someone else's body, die and become a ghost, or turn myself into a pickle, I'd still identify as a man despite not having XY sex chromosomes. — Michael

Taking the first half 'psychological'... If you had never met any other people, would you still identity as a man, and if so, how would you know what the word meant?

He calls it "paternalistic BS" to impose known suffering on someone else just because you think it has a higher meaning. I don't agree with him here, but this is an excellent and critical question I think ought to be adressed — Albero

"X is paternalistic BS" is not a question.

People with strongly individualist, neo-liberal ethics don't want the human race to continue after they're gone. Are we really surprised at all by this?

I don't think it should even be dressed up as serious philosophy. It's just one of the extensions of the modern obsession with the self. "All that matters is my comfort and if I'm in the slightest bit inconvenienced then it's a problem". The "it's not fair to bear children" stuff is just a logical extension of this basic selfish premise, which - given no community ethic - they see no reason to avoid.

All the arguments basically boil down to - no-one should be able to impose anything at all on me, I owe no-one anything etc. The basic individualism which drives modern capitalism.

Do you really think there will be an increase of half a million deaths from TB in the US because of Covid lockdowns? That seems extremely far fetched. — Metaphysician Undercover

The Lancet article was not referring to the US, so no. But for countries like India, I see no reason to disbelieve an article suggesting as such in such an eminent journal as The Lancet. I cited it here as an example of the importance of getting the statistics right ie, not biasing data gathering approaches toward only one outcome of interest.

Do you believe it is accurate to say, according to the conceptual act theory of emotion (the Barrett paper you linked earlier and I reread) that while there are no neural correlates that match the aggregate state of "anger", there are neural correlates that match conceptualising (summarising inferentially) sufficiently similar neural correlates together with "anger"? — fdrake

Yes, so long as we see 'sufficiently' as culturally mediated. I think interoception features are no different to perception features (which is the motivation for Barrett's collaboration with both Friston and Seth), so your list of four individually underdetermining variables is no less appropriate here than it is with perception. The cultural/linguistic priors were the missing element in the effort to map neural states to phenomenological reports.

Even if the role "anger" plays is as a character in a play, that doesn't make it cease existing, it might exist with a changed interpretation (that it's no longer a natural kind with a devoted and human-wide neural mechanism for it, it's instead a contextualised inferential summary for arousal and valence). Ie, there are angers which "anger" marks as a post-processed, publicly accessible, summary. — fdrake

I think it's difficult to for me to make any ontological commitment here. People use the word. The term plays a role, I'm not so sure the thing itself does.

Though that the two things are distortions of the same base image might break the analogy; it could be that the neural correlates of state classes conceptualised as anger wouldn't "feel the same" if you took one process and put it into another brain - the patterns might differ quite a bit over people — fdrake

This is the key to the issue I have with committing to it. I'm not sure though. Do 'games' exist despite only having a family resemblance, or do we only commit to 'activities' with 'games' being a functional term, a speech act which achieves some task in context but doesn't refer? I'm inclined to think of more and more of language as speech acts rather than sortal terms, but there may be limits to this tendency.

Seeing as you already agree with that (I think) and I was addressing a misinterpretation — fdrake

Probably my bad. Yes, I do agree.

If salience and categorisation do influence both dorsal and ventral signals even in the dorsal/ventral severing case, and language can influence both those signals, is the point you're making that language and culture only effect the integration of both streams because they're particularly high order processes? — fdrake

In a sense, although language affect the two stream disproportionately. People with certain types of aphasia have associated difficulties with object recognition, but not with object manipulation. Something might even be used (to an extent) without any apparent ability to recognise what it is. Capgras syndrome doesn't affect fine motor treatment (patients aren't expecting the imposter to have irregular features, or be taller, or something). The models which identify objects 'duck' exist in a quite separate part of the visual processing cascade than those which tell us how to interact with it. There has to be a feedback in a higher order model to say "I reckon it's a duck, check if it's got a bill, duck's have those", than the basic saccade-controlling models which say "there's an edge, they usually end with corners, check along it and see". The former is heavily language and culture dependent, the latter much less so - and all other stages in the hierarchy are somewhere in between in terms of influence.

I'm just going to answer you all at once because you're all saying the same thing "because I feel angry, anger must be a thing (the thing I feel)"

How do you know you feel angry? You know you want to punch something, you know your heart is racing, you know you're inclined to growl, your speech has got louder, you're thinking less rationally... but how do you know that lot is 'anger', or even any specific collection? If you had all the same features but without the racing heart, would you still be 'angry'? How about if we took away the desire to punch something? If we kept the same collection but added a compulsion to laugh out loud, are we still dealing with anger?

From the other side, if I was skipping down the street laughing and singing and said "I'm really angry" would you just accept that, or be inclined to think I'd used the word wrongly. If the former, then how do we learn how to use the word in the first place?

So when you say "I feel angry, so there must be such a thing as 'anger'", what is it you're committing to the existence of? The loose grouping of physiological states? - seemingly not, as this would be just a taxonomic question, we can class things however we please. The physiological states themselves? - again, seemingly not as these are not denied by what I'm saying, and yet there's still problems. Your personal preference for identifying some group of physiological states? - would seem odd to defend what amounts to a private language.

So I'm left no clearer from this latest barage of indignation what exactly you're all insisting exists.


A second issue which seems common to you all is the reach of phenomenal experience. Take an example of referred pain. A sufferer might say "I'm in pain, I think there's something stabbing inside my thigh and it's shooting down my leg". The doctor will carry out a series of examinations. On finding no nerve or tissue damage in the thigh he might think about referred back pain. We take no issue with him saying something like "I know it feels like there's something stabbing inside your thigh, but there isn't, you're mistaken. What's actually happening is that you have some tissue damage in your back". Or, if he finds no damage there he might consider the pain neuropathic, or even (worst case) made up entirely. Either way, we consider his knowledge of physiology to trump our gut feeling about the cause.

Given the trivial normality of such an occurrence, I'm baffled by the resentment that anyone might dare do the exact same thing with mental processes. Why would you assume privileged and accurate access to your mental states when you already know you have no such privilege over your bodily states?

The phenomenological end result is the thing we're taking seriously (the feeling of pain in the thigh, the feeling of anger), not the phenomenological 'gut feeling' about how such a result came about.

So if I say 'anger' is a public model of certain physiological states coupled with socially-mediated contextual parameters from the environment. That is not denying the end result. It's explaining how it came to be. Interoception detects certain physiological states, environmental cues provide context, learned cultural artifacts and language provide options - you land on 'angry' as the best model for the cause of those physiological states. That's an explanation. It doesn't deny anything except your arbitry armchair guesswork as to how your mind works (which I'm not going to apologise for denying).

Translating it back to make sure we're concordant: the priors=flippers, task parameters =pegs and the strength of the trigger = hidden states. — fdrake

Yep, that's it.

I think what I claimed is a bit stronger, it isn't just that the hidden state variables act as a sufficient cause for perceptual features to form (given task parameters and priors), I was also claiming that the value of the hidden states acts as a sufficient cause for the content of those formed perceptual features. So if I touch something at 100 degrees celcius (hidden state value), it will feel hot (content of perceptual feature). — fdrake

Yeah, I'd agree there and I think it's an important distinction because values of hidden states do more than just work for object recognition, objects can be in states.

in order for the perceptual features it forms to be fit for purpose representations of the hidden states, whatever means of representation has to link the hidden state values with the perceptual feature content. If generically/ceteris paribus there failed to be a relationship between the hidden states and perceptual features with that character, perception wouldn't be a pragmatic modelling process. — fdrake

Yes, to a point. I think you and I might agree on this in principle but perhaps disagree in extent. My issue comes when we translate values of objects of perception into words (like temperature) and so in doing model them with a heavy reliance on priors (to make them mutually understandable). The hidden values of hidden states certainly must be relatively faithfully represented by our models in order for them to function pragmatically, but the public models of those hidden values need not. They only need serve their social purpose of inter-translatability. Which is not to remove them entirely from the hidden values - there's nothing quite so useful for inter-translatability as strong reference to a shared variable - but it does reintroduce the effect of socially mediated priors.

Let's take showing someone a picture of a duck. Even if they hadn't seen anything like a duck before, they would be able to demarcate the duck from whatever background it was on and would see roughly the same features; they'd see the wing bits, the bill, the long neck etc. That can be thought of splitting up patterns of (visual?) stimuli into chunks regardless of whether the chunks are named, interpreted, felt about etc. The evidence for that comes in two parts: firstly that the parts of the brain that it is known do abstract language stuff activate later than the object recognition parts that chunk the sensory stimuli up in the first place, and secondly that it would be such an inefficient strategy to require the brain have a unique "duck" category in order to recognise the duck as a distinct feature of the picture. IE, it is implausible that seeing a duck as a duck is required to see the object in the picture that others would see as the duck. — fdrake

Yes, that's how I understand it.

That indicates that the elicited data is averaged and modelled somehow, and what we see - the picture - emerges from that ludicrously complicated series of hidden state data (and priors + task parameters). But what is the duration of a perceptual event of seeing such a face? If it were quicker than it takes to form a brief fixation on the image, we wouldn't see the whole face. Similarly, people forage the face picture for what is expected to be informative new content based on what fixations they've already made - eg if someone sees one eye, they look for another and maybe pass over the nose. So it seems the time period the model is updating, eliciting and promoting new actions in is sufficiently short that it does so within fixations. — fdrake

Yes. We could interpret a single saccade as a perception event if there's a model generated by it. It depends on the hierarchical level we're interested in

But that makes the aggregate perceptual feature of the face no longer neatly correspond to a single "global state"/global update of the model - because from before it is updating at least some parts of it during brief fixations, and the information content of brief fixations are a component part of the aggregate perceptual feature of someone's face. — fdrake

I'm not entirely sure what you mean here.

What that establishes is that salience and ongoing categorisation of sensory stimuli are highly influential in promoting actions during the environmental exploration that generates the stable features of our perception.

So it seems that the temporal ordering of dorsal and ventral signals doesn't block the influence of salience and categorisation on promoting exploratory actions — fdrake

Salience and categorisation effect models present in both pathways. Recall the poor monkeys with thier ventral and dorsal pathways severed. The model they use to handle the banana (knowing his soft it is, how heavy...) is informed by priors from the entire previous experience, but at the time of the perception event (thinking about a cortical level higher than a single saccade here), the two are not connected.

So, in your face example, at this higher cortical level, the movement from one feature to where the next feature is expected might be initiated by the link between the dorsal pathway and the sensorimotor systems which move the eye. These models are already in place in the fusiform gyrus, informed by object recognition models just above them (in the hierarchy).

So it seems that the temporal ordering of dorsal and ventral signals doesn't block the influence of salience and categorisation on promoting exploratory actions; and if they are ordered in that manner within a single update step, that ordering does not necessarily transfer to an ordering on those signal types within a single perceptual event - there can be feedback between them if there are multiple update steps, and feedforwards from previous update steps which indeed have had such cultural influences. — fdrake

That's exactly right, as far as I understand it.

The extent to which language use influences the emerging perceptual landscape will be at least the extent to which language use modifies and shapes the salience and categorisation components that inform the promotion of exploratory behaviours. What goes into that promotion need not be accrued within the perceptual event or a single model update. That dependence on prior and task parameters leaves a lot of room for language use (and other cultural effects) to play a strong role in shaping the emergence of perceptual features. — fdrake

Yeah. Essentially once models are generated in lower level cortices, they need not refer back to the signals which informed the model acting as a constraint on it (at the time of a perception event). But it will have been strongly influenced by them during the course of its formation (prior to the perception event). I know I've said this before, but this is why infant psychology is so fascinating, what we see in infant eye tracking experiments is the actual formation of some of these saccade level models.

I'm going to try and explain the bigger picture which might answer the questions in your subsequent posts. If I can borrow from fdrake's post to start with as he put it very succinctly

(1) Environmental hidden states underdetermine perceptual feature formation.
(2) Bodily hidden states underdetermine perceptual feature formation.
(3) Task parameters underdetermine perceptual feature formation.
(4) Priors underdetermine perceptual feature formation. — fdrake

Those four variables together determine the perceptual feature (or interoceptual feature like emotion). Public models influence priors because in a social group with language there's a significant advantage in maintaining a strong similarity in our priors (language simply wouldn't work without it - thanks @Banno for helping draw this particular thread together for me).

So the extent to which a perceptual feature (an object of conscious experience, if you like - that which we are conscious of) is determined by hidden states is limited. These variables will have different strength depending on the feature. Some might be most heavily influence by hidden sates (catching a ball, for example), others will be more heavily influenced by prior (emotions - see the Barrett paper I cited earlier). None can be fully determined by anything less than all four, and in normal circumstances all are fully determined by all four (there are no other factors).

Translating a conscious experience into words or social relevant actions (which is most of them) absolutely requires that task parameters and priors strongly dominate the modelling because they are the only areas where it is possible to maintain inter-translatability, and without that the whole use of language or social-relevant action would be unlikely to succeed)

So when modelling my interceptive physiological signals in a task where a socially-relevant action is expected (or likely), I'm going to be reaching for models which are more publicly shaped so that my responses are more likely to have the expected result.


To your specific questions, with that in mind...

How does that work for animals? Fear and aggression are important for survival, and they're not exactly querying themselves for reports on conscious experiences. — Marchesk

Animals need to model the causes of their internal states no less than we do to help make predicatively useful responses, and if those responses are socially-relevant, they will need to be modelled on something public. Obviously it would not be advanced language in this case, but observed behaviour.

In animals without the mental structures to create modelling hierarchies, I can see why certain responses are important for survival, and certain physiological states which prepare for or facilitate those responses, but whilst useful, I can't see more advanced modelling like emotions being necessary for survival.

Also in the moment when someone punches me, I'm probably reacting in anger, not stopping to do some reflection. That comes after the reaction. — Marchesk

Well, you'd have to do some research work to establish that wouldn't you? Which is what Barrett and Seth (and others) have done. The results are that no, you do not react in anger, because there is no neural correlate for anger.

You earlier said

I fully endorse what Dr. Seth is doing. — Marchesk

We're not going to get anywhere in his stated project if you just insist that whatever you feel is happening is what's actually happening. In order to map brain states to phenomenological reports, you have to accept what neuroscientists are saying about brain states, otherwise you're saying phenomenological reports are infallible in a way the neuroscience isn't. So neuroscience says there is not an identifiable neural correlate for anger - we've looked really hard and can't find one. You've two choices 1) insist that because it feels like there must be one to you then that's the case and neuroscience just isn't trying hard enough, or 2) accept that something feeling like it's the case is not necessarily proof that it is, in fact, the case and work out how those feelings might have come about.

As I said to Khaled, if you're of the former persuasion, there's no point in us talking (there's no point in talking to anyone). If you're just going to assume that the way things seem to you to be is the way they actually are regardless of any evidence to the contrary, then there's no point in seeking other views is there?

So does this mean other animals do not have experiences of colors, tastes, memories, because they lack the language to ask themselves about how other animals typically react? — Marchesk

Not necessarily. Public models don't have to be linguistic, they can be behavioural. All public models do is influence priors, anything influencing a prior as a result of cultural homogenisation is a public model.

Thinking about this some more, how would the words "afraid", "red" or "pain" have become part of language if there wasn't fear, color, or uncomfortable sensations to begin with? What exactly is the public model that we learn based on? — Marchesk

I'm no linguist, but the history of the origin of words is hardly a simple matter of us finding new things to name in the world is it? Words are used. It wasn't for idle fun that @Banno made this point earlier. It's quite important. Words don't need to refer, so the mere existence of a word does not imply that some exact referent of it exists. My guess is that if you look to the use of these words you'll find how and why they came about. Why do I tell someone I'm in pain? Maybe to get help, maybe to get them to stop doing something, maybe to feel part of a social group (where being in 'pain' might be a membership criteria). any one of these uses might be why the word came about. But none of these speech acts are going to work unless I can model the physiological signals I want to attend to in a way that is translatable to the people I want to do something for me. That's where the public models come in.

What matters is that phenomenological experiences exist and need to be accounted for. We see colors. We feel emotions, pains, taste food. We dream. We visualize. Many of us have inner dialog. We relive memories at times. — Marchesk

No one's denying any of that from a phenomenological perspective. It's just that from a process perspective some of those accounts are not as we think they are. When you feel 'angry' it feels like you're 'finding' yourself to be in some state, but you're not. There's no such state. It doesn't exist. So that can't be right, no matter how much it feels like it is - or else we discard the idea that conscious experience is caused by the brain, in which case why bother looking at it at all.

So people like Barrett try to find out what's going on. How can a set of physiological states with no boundary and no non-overlapping properties give rise to the feeling that we're 'angry'? The answer she proposes (and with substantial empirical support) is that we use public models to infer the causes of our interocepted signals. "I've just had someone punch me, people get 'angry' when they're punched, these mental states I'm receiving data about must be 'anger'"

Same can be said of colour, tastes, memories... the more we look, the more useful an explanation this model provides.

I don't know wha it means to say fear is a public model. I can't always tell when someone is afraid. — Marchesk

It's not about you telling if someone else is afraid. It's about them deciding that they themselves are afraid.

The article was concerned with "US Covid-19 Death Counts". You flippantly mention "nearly half a million excess deaths" — Metaphysician Undercover

How could you possibly consider citing an article from The Lancet warning of nearly half a million deaths from TB as a result of Covid strategies whilst we were talking about the statistics used to determine that very strategy 'flippant'. What on earth was 'flippant' about it?

it sure as hell seemed like you were arguing along eliminativist lines to me and others in this thread. In fact, in the very post before your reply to me you're doing it again. Replacing the experience of fear with talk of a model and public convention. — Marchesk

One of the possible mappings of brain activities to phenomenological experience is via public models like 'fear'. Why would you rule that out?

I wonder why that is. :chin: — Marchesk

Because brains are just lumps of biological matter with electrical and chemical activity. Just looking at it isn't going to tell us what any of it's doing any more than looking at a microprocessor is going to tell us what software is on it.

The point is that once certain mappings have been established (and a huge quantity have), then we get correlations, strongly predictive models, statistical inferences... Once we reach a certain threshold we can start to look at aberrant phenomenological reports and say "well, either this report's not quite right or we have to throw away all these otherwise excellent models". Since we've absolutely no reason to presume phenomenological reports are always accurate (as in providing the type of data point we're interested in), we don't just take them at face value anymore as we might have done at the outset of the project, hence Seth and Barrett's work on public models of phenomenon like emotions and colour.

Methodologically, yes. I wouldn't have put it that way (I don't literally agree with every word he says, of course), but broadly speaking yes.

In psychology there's very little choice but to start out with self-reports and ask "what's going on to cause this?" We can't just look at brains and expect to 'see' what's going on without any phenomenological data. We get people to say what they're experiencing, we look at their behaviours, we correlate these with brain activity (and other behaviours) and make inferences.

I don't know what to say. We could trade Anil Seth quotes all day. I've read most of his papers (certainly the Sackler lab work, anyway). I've been to three of his lectures and I've worked, briefly, with a couple of people from his lab. You've either misunderstood his position or you've misunderstood mine because a large chunk of my position on this comes from Seth's work. He's certainly not opposed to the position I'm expounding here and he's not a supporter of Chalmers, Nagel et al's position on this.

It's not that I've independently come up with a theory and now I'm saying "look Anil Seth agrees with me", it's mostly his theory that I'm presenting here. Him, Friston, Barrett, Edelman... all of whom frequently collaborate on the same papers and are very close in their view on this.

Nothing prevents them from saying it. But they haven't had said experience. Therefore they do not know what they're talking about. — khaled

I haven't had a rollercoaster ride, but I know what one is. I can use the term correctly. I don't see how you can justify a difference with 'fear'. Why do I need to have experienced 'fear' to know what I'm talking about when I use the term, but I don't need to have experienced a rollercoaster ride to know what I'm talking about when I use the term?

The contradiction is you saying that fear is a public concept and not an experience and at the same time that fear is an experience. — khaled

Experiences themselves (as a models of interocepted states) are public concepts. That's what the Barrett paper was about. Do you read the stuff I cite or not, because it's not worth my while doing so if not?

Fear is further a category of experiences. What belongs in that category is a public convention.

For instance, if Helen Keller never learned to communicate with people, I would still assume she was conscious. — khaled

My bad, we've had this misunderstanding before and I haven't learnt from it. 'Reporting' has a technical meaning in cognitive science, it doesn't necessarily mean spoken or written. Think of it as a writing a journal in your head.

That's what being scared is, not what it's like. — Isaac


So being scared is the experience you have on a horror ride? You're contradicting yourself:

First you insist that if someone who has never experienced fear before (someone with urbach-wiethe disease even) uses the word "afraid" then they know what fear is. Now you say that fear is fundamentally an experience. — khaled

What is preventing someone with urbach-wiethe disease (passing over the complications in simply correlating the condition with a lack of ability to feel fear) from saying "being scared is the experience you have on a horror ride". If being scared is the experience you have on a horror ride, then someone correctly identifying it as such has understood what fear is, haven't they? I don't see the contradiction.

You can be unable to report on working memory and still have experiences. — khaled

How could you possibly know that?

Thanks. That pretty much ties in with my understanding of Seth's position from his papers. The aspects I don't see how you're attributing are things like...

when they do cover the statistical inference of perception, conscious experience is still the end result of that which needs to be explained. — Marchesk

...and...

he does not dismiss phenomonlogy by replacing with with neurological or statistical terms, as you do. — Marchesk

Seth's work, his research objective in fact, is to do exactly that, explain the one in terms of the other. When he talks about matching first-person reports to third person analysis, he's explaining his method, not reifying first-person reports.

Neural correlates of consciousness wouldn't make sense unless Seth (along with Crick and Koch) didn't take phenomenal consciousness seriously as something in need of explanation. — Marchesk

Again, methodologically, not ontologically.

This last quote from the paper is exactly what the anti-Dennett side has been arguing this entire thread.

But as powerful as these experiments are, they do not really address the ‘real’ problem of consciousness. To say that a posterior cortical ‘hot-spot’ (for instance) is reliably activated during conscious perception does not explain why activity in that region should be associated with consciousness. — Marchesk

No, you've misunderstood what he's saying here. He's saying that the posterior cortical activity could not explain why the region should be associated with consciousness, not because of some fundamental inability to provide such explanations, but because the specific functions within that region don't encompass a wide enough base of signals related to conscious reports. He's making a purely neurological point, not a deep philosophical one. He's just saying 'good as these single modality correlations are, the don't address the real problem because it is multi-modal. He's not saying anything like what's being advanced on this thread. His entire lab would be rendered pointless if he held to that view.

Seth's View on perception is basically where I'm getting a lot of what I'm saying here.

https://www.tandfonline.com/doi/abs/10.1080/17588928.2015.1026888, is unfortunately paywalled, but this paper gives a reasonably good account. He talks specifically about an active inference model of colour synthesis explaining conscious perception of colour.

Perception of a tomato, on this view, involves the brain deploying a high-level generative model predicting the sensory responses elicited by the tomato. In contrast to sensorimotor theory, PP emphasizes neural mechanisms as both necessary and sufficient for perceptual experience (at least at any particular instant)...

..My specific claim is that the subjective veridicality (or perceptual presence) of normal perception depends precisely on the counterfactual richness of the corresponding generative models...

...In addition to accounting for the phenomenology of synesthesia, the theory naturally accommodates phenomenological differences between a range of experiential states including dreaming, hallucination, and the like.

[my bolding]

I wouldn't. I would say "It's what you feel when you go on a horror ride" and ask you to try it. — khaled

That's what being scared is, not what it's like.

How do you know rocks don't have a mental state? — khaled

It's not the mental state, it's the inability to report on working memory, which you'd just said was what 'experiences' are. Rocks don't have a working memory.

It's just that we only know that a human's neural network produces consciousness. And an AI is fundamentally different in that it doesn't have neurons. They are not similar enough to conclude both are consciuos. — khaled

This assumes consciousness is very tightly bound the the type of substrate. I'm not even sure I'd go that far.

you can't guess what fear feels like from a first person view. This "what fear feels like" is qualia. — khaled

What would an answer to this question even be? As far as I can tell it doesn't make any sense at all. If I ask "what the the rollercoaster like? " you might say "it was scary". If I ask "what was being scared like?", I expect you to shake your head and walk away, what could I possibly mean by that?

Simply being able to form a sentence does not make the content meaningful.

we just have this one experience of what's going on right now. — khaled

But you don't. That's the point. You have a memory of what was going on a few seconds ago. There's a fundamental disconnect between the external world (if you believe in such a thing) and your experiences which makes talk of the experience of red - where 'red' is considered to be something in the external world) fundamentally wrong.

If you want to say that experience 'just is' the unified memories of some mental states from the last few seconds, then we can work with that, but then we have an very good model of that already. There's no need for qualia.

panpsychism is wrong, and physicalism is fine — Isaac


Don't see how either of those follow. — khaled

Because if conscious experience is just reaching for some word (or other response) from some internal mental state, then rocks can't do it and we've given an entirely complete physical account of it.

If it is then AI is definitely conscious because it can reach fr the word 'red' in response to some state of it's neural network. — Isaac


That doesn't follow exactly. An AI's "neural network" is hardly similar to a human's as far as I know. — khaled

What part of the definition of conscious requires that is takes place in a network similar to humans?

You'll have to quote him (or we'll just agree to differ), it's not the impression I get from either that podcast, nor his other lectures, nor his papers.

And you confirmed that you also believed, as I do, that the reporting practice was not as stated in the article, — Metaphysician Undercover

The reporting practice was exactly as stated in the article.

concerns [were] raised by academics from the Centre for Evidence-Based Medicine about the original measure, which counted anyone who had ever tested positive as a COVID-associated death.

I don't understand how much more clear that can be. It says they used to count anyone who ever tested positive as a Covid-associated death, which is exactly what the article claims. And unsurprisingly, since rather than just writing about whet he reckons might be the case, the author has just quoted the two sources from which he gained the information.

OK, so at the beginning of the pandemic, when there was only a few people in the population of a given country who tested positive already, there was even less people who tested positive and died. The practice of counting everyone who tests positive and dies would produce a very small number of mistakes, even if it might have been a somewhat significant percentage of the overall count, at that time. — Metaphysician Undercover

How many cases and deaths were there at the time of the change?

relative to the overall numbers, the mistakes reported at the beginning when there was a very small number, are very insignificant, constituting a very small percentage of the overall numbers. — Metaphysician Undercover

What percentages of the deaths recorded were recorded under the old/new system?

By "tricky complications" you really mean deceptive speak. — Metaphysician Undercover

If you're going to be that flippant about nearly half a million excess deaths then there's no point talking to you. I don't suppose you give a shit about what we do if you don't even care how many people die. It's pointless trying to talk to someone about strategy whose motives I don't share at all.

What evidence do you have that that's what you did? You learnt to use 'red' at, what, two, three? Are you suggesting you have a clear memory of the method you used? — Isaac


Be reasonable. What use is it asking the question if the reply was going to be: Actually, you don't remember. — khaled

Because your own ad hoc, single person sample is next to useless as a description of how people learn to use terms. I thought you might have, you know, read some actual research before just randomly deciding how the cognitive development of language works, which you might be able to point me to, failing that some uniquely clear memory to work with, As it turns out, it's just how you 'reckon' it probably works based on the five minutes thought you gave it just now. There's masses and masses of research time gone into trying to figure this stuff out, you know. I know my tone can sometimes get a bit short, but can you not see how frustrating it is to be in a position where it might take years of work and ruthless scrutiny from peers to get to a point where I could publish a paper just on one small aspect of how our mind works only to have a discussion touching on those facts dominated by a load of guesswork about it from an armchair after not even having read the results of such investigations let alone bothered to carry any out.

Basic algebra tells you that X can take on any value including Y or Z. Point is that it seemed like something. I later call it "red" or "pain" or whatever. — khaled

No. I'm arguing here about the privacy and accesibility aspects. That requires that when you say it seems like X you're right - ie it could not be the alternatives Y and Z.

I'm trying to argue that they are not as you, seconds later, think they were. — Isaac


Agreed. — khaled

Right. So your account of the fact that it 'felt like X' is no more accurate than my neurological account of how it probably felt. We're both guessing how it felt from evidence - mine neurological (statistical likelihoods), your is inferential (traces of working memory re-firing of neurons). Neither have good access, neither have private access.

As far as I can tell, the working memory and sensory memory are the source of experiences. As in if they stopped funcitoning, you wouldn't have any experiences at all. What you're saying here is that I had the experience Y first which was then altered to a different experience X due to built in inaccuracies. That doesn't make sense, what is this experience Y? All I ever see is the experience X. There is no "more accurate" experience Y that preceded it. — khaled

Which is it, the working memory or the sensory memory. It can't be both, they'd deliver contradictory experiences?

Again, equivocation on 'experience' here is causing problems. If you're saying that the working memory is the 'source' of experience - ie it generates, but does not constitute experience - then that's a whole different discussion than the one we're having about perception, which involves considerably more brain regions than signal to the working memory. It would help if you clarified what model of consciousness you were working from.

If I am measuring something and it turns out to be 5cm you cannot make the claim "Actually, you made a more accurate measurement which was then changed to 5cm +- 0.1cm due to the built in inaccuracy of the ruler". — khaled

Yes I can, if I've got good evidence that that's what's happening. Why would I not?

Conscious experience is invoked in AI, physicalism, the limits of knowledge... — Isaac


Can't AI also have a certain experience then reach for the word "red" to describe it? — khaled

Yes, I think it can. I was pointing out here that theories about consciousness matter - in opposition to your comment about 'experience' just being shorthand for this. If it is then AI is definitely conscious because it can reach fr the word 'red' in response to some state of it's neural network. Yet there's intense debate about whether AI is conscious or could ever be. So this equivocation isn't helping. It' not the case that 'qualia' is just shorthand for experience which is just shorthand for this correlation between mental state and tendencies to respond (like reaching for the word 'red').

'Experience' is being used to refer to some ineffable, private, introspectively accessible concept when it come to AI, p-zombies, etc. Then when pushed by things like Dennett's intuition pumps and the evidence from neuroscience, you retreat to just "whatever you just described - that's what we mean by 'qualia'". But then the questions drop away. Ai is already conscious, p-zombies are impossible, panpsychism is wrong, and physicalism is fine - job done.

At no point do I have a 'feeling of a colour' which I then select the name for from some internal pantone chart. — Isaac


But you said that you experience something, then reach for the word "red" to describe it. I am asking how we can compare these "somethings". — khaled

Reaching for the word 'red' is part of the experience. As @creativesoul has pointed out experience is a constant process, not a series of discreet packages.

But the article says it would be recorded as a Covid-19 death even if Covid-19 wasn't an exacerbating factor. That's a big difference. — Metaphysician Undercover

Yes. Why would you just prima facie disbelieve this? He's provided sources, and the was good reason to, as John Newton at PHE said

The way we count deaths in people with COVID-19 in England was originally chosen to avoid underestimating deaths caused by the virus in the early stages of the pandemic.

Since then changes have been made to include epidemiological evidence that Covid -19 was indeed an exacerbating factor.
I don't know as much about the situation in America, but I see no reason why it wouldn't have been the same. It was a good decision.
From the UK government's own website

concerns [were] raised by academics from the Centre for Evidence-Based Medicine about the original measure, which counted anyone who had ever tested positive as a COVID-associated death.

the numbers of deaths in people who have tested positive have become substantially greater than the numbers of deaths subsequently registered as COVID-19 deaths by the ONS, which is why we are now changing our approach to reporting deaths — PHE

What's interesting here is not the facts themselves, which are as indisputable as it gets, but the way in which, without even researching the article's sources, you've already assume it is

exaggerated, deceptive, and arguably false — Metaphysician Undercover

It's pretty much uncontested that policies designed to reduce deaths from covid will cause a rise in deaths from other causes. — Isaac


Actually I think this is a very dubious statement. — Metaphysician Undercover

These figures might support a contention that the measures being taken to deal with covid-19 may be having a negative effect on other causes of death, particularly other respiratory causes — BMJ

Since the week ending June 26 there have been more non-coronavirus deaths registered above what would usually be expected in private homes than deaths registered involving Covid-19 — Office for National Statistics

...which is just to address the direct correlation - not even touch on the knock on effects...

Even temporary disruptions can cause long-term increases in TB incidence and mortality. If lockdown-related disruptions cause a temporary 50% reduction in TB transmission, we estimated that a 3-month suspension of TB services, followed by 10 months to restore to normal, would cause, over the next 5 years, an additional 1⋅19 million TB cases (Crl 1⋅06–1⋅33) and 361,000 TB deaths (CrI 333–394 thousand) in India, 24,700 (16,100–44,700) TB cases and 12,500 deaths (8.8–17.8 thousand) in Kenya, and 4,350 (826–6,540) cases and 1,340 deaths (815–1,980) in Ukraine. The principal driver of these adverse impacts is the accumulation of undetected TB during a lockdown. — The potential impact of the COVID-19 pandemic on the tuberculosis epidemic a modelling analysis - The Lancet

... but you carry on with your preferred narrative, don't let any of these tricky complications get in the way of your all-American hero flick.

So they're "mutually exclusive" in terms of being qualitatively distinct variable types in the variable network of the model, but they're not thereby causally or statistically independent of each other since they're connected. — fdrake

Cool, that makes sense. We have common ground here.

It isn't as if the hidden states are inputs into a priorless, languageless, taskless system, the data streams coming out of the hidden state are incorporated into our mature perceptual models. In that respect, it does seem appropriate to say that the hidden states do cause someone to see a rabbit or a duck, as one has fixed the status of the whole model prior to looking at the picture. — fdrake

I see what you mean here. If at any given time the only variable that really is 'varying' in the system is the hidden state, then we can appropriately talk about a direct causal relationship. Like triggering a pinball, the various flippers and pegs are going to be determinate of it's path, but they're fixed, so right now it's path is directly caused by the strength of the trigger?

So, if we want to answer the question "what are people modelling?" I think the only answer can be 'hidden states', if they were any less than that then the whole inference model wouldn't make any sense. No-one 'models' and apple - it's already an apple.

But...

If we're talking about the properties of those hidden sates which constrain the model choices...

So the issue of the degree of "fuzziness" associated with labelling hidden state patterns with perceptual feature names comes down to the tightness of the constraint the hidden states place upon the space of perceptual features consistent with it and the nature of those constraints more generally. — fdrake

I'd agree here. Do you recall our conversation about how the two pathways of perception interact - the 'what' and the 'how' of active inference? I think there's a necessary link between the two, but not at an individual neurological level, rather at a cultural sociological level. All object recognition is culturally mediated to an extent, but that cultural categorising is limited - it has functional constraints. So whilst I don't see anything ontological in hidden states which draws a line between the rabbit and the bit of sky next to it, an object recognition model which treated that particular combination of states as a single object simply wouldn't work, it would be impossible to keep track of it. In that sense, I agree that properties of the hidden sates have (given our biological and cultural practices) constrained the choices of public model formation. Basically, because the dorsal pathways activities in object manipulation etc will eventually constrain the ventral pathways choices in object recognition, but there isn't (as far as we know) a neurological mechanism for them to do so at the time (ie in a single perception event).

A little of what we know. Object recognition in adults is mediated by two systems. A high level one which relates to naming, conceptual properties (such as use, ownership etc) and manipulation. But we also had what's called a mid-level system which is responsible for object tracking and enumeration. This system appears (in adults) to be independent (broadly meaning lesions in it can cause independent issues). Here spatiotemporal signal are king (what moved relative to what), things like edge, colour, shape etc play a secondary role in the case of stationary objects. But none of this directly informs the higher levels system (at the time of a single perception event). The higher level system is extremely culturally mediated, and is very difficult to actually change by perceptual features alone.

Experiments on surprise and attention in infants have indicated that their own object enumeration relies heavily on spatiotemporal markers and so it seem likely that this system is the primary object division system and the higher-level one is secondary. Interestingly, infants as young as 2 months show strong object recognition in this primary mid-level system, but not until 18-24 months do they have an equivalent grasp of object recognition in the higher system.

So higher level it might first go... hidden state properties > some constrained model space > cultural/biological modelling process > object christening

Then in an instance of perception... hidden state properties > some (now constrained) choice of public models > object recognition

But mid-level it would go hidden state spatiotemporal properties > (possibly drawing on other mid level properties - shape, edge etc) > object enumeration > object recognition

In order to have this model we have to have a cognitive facility to model new data (cultures are groups of people after all), so in that sense I agree that

language use plays some role in perceptual feature formation - but clearly it doesn't have to matter in all people at all times, just that it does seem to matter in sufficiently mature people. — fdrake

We may disagree as to the extents, but I think we have common ground on the general process.

in your case, you interpret pretty much everything in terms of a scientific framework - objective facts, satisfactory explanations and so on. That is internalised in such a way that it becomes second nature to you. — Wayfarer

...and you know this how?

It’s about the logical contradictions of materialism. Logic is important for some. — Olivier5

And you think those taking an alternative position to you don't think they're being entirely logically consistent? So you think their position has logical contradictions...they obviously don't. What now? You point out the logical contradictions, they say "no, they're not logical contradictions because...".

This is exactly the discussion that's been happening in the field for hundreds of years. The discussion that Dennett and Chalmers et al are still having, that their respective works are part of. It's not only naive, but unbelievable arrogant to think you're the first one to suggest we use logical contradiction to analyse the positions. It's not as if either side have just written a three line syllogism that can be just put into a truth table or something. Even just parsing the two arguments into formal logic would be fiendishly difficult and prone to error, let alone the task of then comparing the two for logical errors

But then by 'logic' you don't really mean Logic do you? As with most people like you here, when you use 'logic' you just mean 'what seems to me to be the case'.

the explanations are just replacing phenomenological terms with statistical ones. That's not an explanation. It's equivocation. — Marchesk

I don't understand what you mean by this. Perhaps you could clarify with some examples from Seth's papers?

Ok. Everyone hang their gloves up, fight's over. Someone thinks it's silly, so that's settled the matter to everyone's satisfaction...

...but hang on, some other writer doesn't think it's silly...oh no, now what will we do?

It's almost as if it's a difficult issue on which many intelligent people have differing opinions...

But by all means just find whichever one agrees with you and cite them as if they were gospel.

Yeah, but he doesn't dismiss the problem as just a philosophical misuse of language. — Marchesk

Actually he does (to an extent). I'm fairly certain he used almost those exact words in a lecture. I'll see if I can find something more concrete for you, so that you don't just have to take my word for it, but for now, don't confuse his tackling the problem with his not seeing a terrible linguistic muddle also. He's not a philosopher of language and doesn't make a habit of making propositions outside of his domain, but I've attended a few of his lectures now and the linguistic issues are not lost on him.

it's a topic for neuroscience to resolve. I'm open to that if it actually explains how colors and pains arise from brain processes. — Marchesk

Well, that depends on what you mean by 'explain'. A problem I find with many 'consciousness' arguments. What serves as an explanation is very subjective. One can obviously continue to ask '...but why?' ad infinitum, so when to stop doing so is a personal choice.

That said you seem a little resistant to the ideas I've already written about on this (not my ideas of course, I'm just regurgitating). If you're open to neuroscience explaining these things then whence the resistance? Are there some explanations you find particularly unpalatable?

But one could say the same thing for using words like model for sensation. — Marchesk

I don't think so. The idea of sensation being filtered through Bayesian models is expounded in great detail in the various papers on the subject. Not everyone agrees that it's a good or even accurate way of modelling cognition, but I haven't read anyone suggest there's a problem with equivocation on terms. I can't even think what that might consist of, did you have something specific in mind?

I did start a thread a year or so ago where neuroscientists Anil Seth discussed in a podcast his research into consciousness and marking progress on the hard problem. — Marchesk

Cool, I'll have a read sometime.

So not just a few misguided philosophers. — Marchesk

You know Anil has categorically said there's no hard problem of consciousness, right? You've possibly misunderstood his line of research. He's attempting to answer that very question using neuroscience - specifically a Bayesian inference model. That means he believes a) it's possible to explain phenomenological experience using neuroscience, b) the cause of phenomenological experience is not introspectively available, and c) that phenomenological experience can be studied third party.

I'm not sure how that sets him in the same gang as people like Chalmers (whom he's openly said is wrong about consciousness).

Read some of Seth's papers, he's a lot less circumspect than he is in public lectures.

As to how I learned, I looked at all the situations where people said "red" and found out the common factor in my experience, that is "redness". — khaled

What evidence do you have that that's what you did? You learnt to use 'red' at, what, two, three? Are you suggesting you have a clear memory of the method you used?

I claimed that that I am experiencing things in the first place is an unarguable fact. — khaled

You said...

I want to emphasize that the statement "the world seems like X to me" is not negated by any neurological evidence you can throw at it. The world still seems the way it seems. — khaled

You didn't say 'the world seems like something'. You said ''...seems like X". I'm saying, for example, that the evidence from cognitive science suggests that it cannot have seemed like X. It must have seemed like Y, or Z. You're simply reporting, post hoc, that it seemed like X because of your cultural models which encourage you to talk about experiences in this way.

No-one is denying you have experiences. I'm trying to argue that they are not as you, seconds later, think they were.

as we decrease the time frame the inaccuracies decrease as well. — khaled

In long term memory, yes. I'm talking about sensory and working memories here. They don't work the same way, the inaccuracies are built in to the mechanism, it happens instantly, as a result of hippocampus function, not long term as a result of action potential changes.

Because you claim at the same time that we have experiences which "we later reach for the word 'red' to describe". People say "we have experiences of colours" as a shorthand for that. — khaled

No, they don't just use it as shorthand. Conscious experience is invoked in AI, physicalism, the limits of knowledge... This is exactly the eqivocation I referred to. You make specific claim about the nature of 'experiences', and then, when pushed on them, revert to "oh it's all just another way of saying exactly what you just said". Having an experience of something and, as Parr of that experience, reaching for the word 'red' is not the same as having an experience of colour. The two have radically different implications.

So if, hypothetically, we could take a screen shot of what I'm seeing and show it to you, how big of a difference do you think can exist? Can you imagine a situation where you remark: "Why is the sky red?" — khaled

No, it's absolutely impossible, thats just not how colour and language is processed in the brain (the link between photon hitting the retina and vocal muscles making the word 'red'). At no point do I have a 'feeling of a colour' which I then select the name for from some internal pantone chart.

What new information is being learned? — Isaac


Which group each belongs to for one. How they're related. — khaled

But each group belongs to whatever category you feel like it belongs to, and they're related in whatever way you feel like they're related, either one of which might change from one second to the next.

Qualia is experience, or an aspect of experience. — frank

'Experience' is no less slippery a term unless pinned down. Equivocation is the weapon of choice for most woo-merchants.

The person died "in" a car crash, not in a hospital. — Metaphysician Undercover

There's a post mortem which will still try to establish the cause of death. If the person in the car died from their injuries at the scene and one excaserbating factor was a covid infection then it would be listed as a covid death. We're talking unlikely circumstances at this stage, but it would be recorded the way the article describes, and for good reason too.

The article doesn't mention any judgement of a "chain of events". — Metaphysician Undercover

It's how death certificates work.

Did you read the article? It seems to have been written with a very bias slant, to me. The way they suggest that Covid deaths ought to be recounted to exclude a whole bunch as illegitimate seems very similar to the way that Trump suggests votes ought to be recounted. — Metaphysician Undercover

In that they're both asking for a recount? Surely the significant factor in Trump's actions is that he's asking for legitimate votes to be discounted. The legitimacy of the picture presented by the statistics for the job at hand is what matters, not the superficial resemblance anyone making such a request shares.

No one is "shutting down any discussion". — Metaphysician Undercover

I was referring here to the general trend, not that specific article.

the thing which Trump is complaining about, a presidential election, seems to be a lot more important than the other thing, number of Covid deaths, which is just statistics used for models. — Metaphysician Undercover

Those models determine policy and public response which, in the current state of crisis, determines who and how many die. So I think they are extremely important.

This does not even address the authors claim of "90% or more effective false positives" in "various types" of testing. I don't know which agencies would be using different types of testing which are known to give results with more than ninety percent of the positives being false positives. — Metaphysician Undercover

No, I don't either, and that claim about false positives should have been supported with a citation or quote from an expert. It doesn't invalidate the consequences the author highlights on the counting of non-covid excess deaths.

It's pretty much uncontested that policies designed to reduce deaths from covid will cause a rise in deaths from other causes. So how we monitor and predict those collateral deaths really matters.

Or we could say "whatever is actually happening in our conscious process, we'll call that 'mind' and work out what properties it has" etc. — bongo fury

Yeah...presuming we have any more confidence that we know what 'our conscious process' distinguishes than we do what 'our mental processes' does.

describes the suspension of judgment about the the objects of experience so as to develop a detached awareness of the nature of immediate experience.

As Frank points out above, the 'raw' nature of experience is generally straighaway incorporated into 'stories' which attempts to situate it in so-called 'objective' terms. We generally do that instinctively, immediately, without noticing. The point of the phenomenological suspension is to notice that. — Wayfarer

I appreciate the explanation, but I'm still not seeing the 'study'. If one performs this 'bracketing' then one has list of experiences which one just accepts unquestioningly as being what they are. Great. What have we learned that we didn't previously know?

I think you're interpreting ”qualia" as "sensory data." — frank

If I was, the process would be unproblematic.

A quale is an instance of a type of consciousness. "Instance" connotes an event here. As Luke put it, it's the end product, which is seamless and unified. That is what we mean by "qualia". — frank

Then how can we have a 'red' quale? Red is not the end result of any stimuli at all. If qualia are now being reduced to just another word for experience where we mean just the recollection of mental states, then it's a) useless, we already have a word, and b)very confusing because there's already a word 'qualia' which is used to talk about subsets of perception (like 'red').

Privacy is just related to the idea that people aren't telepathic. Obviously, in a non-woo sense, we are. I'm trying to read your mind now. The technology I'm using is the written word. So here the discussion would pass into the topic of meaning and truth. — frank

I don't think it's that simple. I think privacy is at the heart of the irreducibility claim, which is far more important for the extent to which neuroscience can investigate. Nonetheless, if I'm wrong, I still struggle to see what properties qualia do have, if not those listed by Dennett.

Incorrect. I do not know where you get that impression. — khaled

Because otherwise you have an experience 'which is red' (Cartesian theatre gone mad), or an experience 'which you called red' (non-private, you can misuse the word). I'm taking the most charitable interpretation of what might be meant by 'an experience of redness', which would be something like - there was some redness>I experienced it. If instead you want to say "the experience I just had is called 'redness'", then I don't know how you'd ever come to learn the word.

Basically, being least charitable "an experience of redness" doesn't make any sense at all. I'm trying to work with a meaning which at least makes sense.

Cool. Has nothing to say about whether or not we have experiences (as usual). — khaled

Yep, because no-one's denying that (as usual).

The timescale issue amounts to "Things are not how you remember them to be or exactly how you describe them to be". This is not an issue of the model. The model is fine, all you have said is that when trying to report this last step (qualia) we give inaccurate reports. I think everyone here already knew that. — khaled

You said the the way things seem to you is a fact and that some philosophical work can be done with that fact. Well it can't. The way things seem to you (as such a fact is available to form part of any philosophical investigation) is not an unarguable fact. The moment you enter it into discourse or consideration it is already wrong, not how things actually did seem to you.

If indeed "everyone here already knew that", then no-one can claim to be having an experience of redness with any more authority than I can claim you're not. You are no more accurate a reporter of the way an event actually felt than I am.

can you imagine a robot that acts identically to a human but doesn't have these "experiences" — khaled

It depends entirely on what you mean by identical. And before you're tempted to say 'exactly identical', have a glance at Wittgenstein on what we could possibly mean by 'exactly'.

Well you seemed to be denying for the longest time. What with "You don't see colors" and all. — khaled

I really can't see why people are finding it so hard to tell the difference between "we don't have experiences" and "we don't have experiences of colours".

what they are experiences of — Isaac


I'm not sure what this question means. — khaled

Do you experience a red cup, or 'redness' and 'cupness', or the mental activities resulting from external stimuli (presumed to be a red cup), or something else? What is the subject matter of this experience.

When I use the term 'experience', I'm just meaning the recollected results of introspection about an event I was just involved in.

we do not know that the same experiences are caused by everyone's brains. — khaled

Agreed, to a certain level of accuracy.

I don't know if when I look at a red apple and you look at a red apple we both have the same expereince. — khaled

Unlikely. Again, depending entirely on the accuracy required.

I know we both call it "red" and it has largely the same relationship in our brains. As in mostly everything I call red you also call red or orange or something around there (assuming neither is colorblind). That does not give evidence that we are experiencing the same thing. — khaled

...and here we go with the 'red' nonsense again. We were talking about experiences - whole events. You don't experience red. You can't it's neurologically impossible. And, as we've just established, you telling me you do has no validity because we've all just agreed that you cannot give an accurate account of you experiences.


As to phenomenology, nothing in that section tells me what it's studying. It says nothing more than "make a list of all the things you think you felt and sort them into groups". What new information is being learned?

"Phenomenology is the study of structures of consciousness as experienced from the first-person point of view" — khaled

So. What's the difference between a 'study' and a 'report'?
One would expect a 'study' of baking a cake to have some kind of hypothesis in mind, data, conclusion - something of that sort. One learns something new from it.
A 'report' might simply be "I dropped the mixture on the floor". One learns nothing new from it, it's merely a conversion of what you already knew to written (or spoken) form.

To claim that phenomenology is the 'study' of structures of consciousness as experienced from the first-person point of view, rather than merely the report of them you'd need to be able to learn something new from it. But if you can't possibly be wrong about what the structures of consciousness are from this perspective (they are exactly how they seem to you to be), then how is it a 'study' and not a mere 'report'?

The word itself seems to presume consciousness. — Coben

Yeah, it's the difference between consciousness as a subject of investigation and consciousness as an adjunct to investigation.

So an illusion is something which appears to be one thing, but is, in fact, another. So immediately it's about the properties of the causes of the perception.

We could say "whatever is actually happening in our mental process, we'll call that 'consciousness' and work out what properties it has - I think that's the route you're drawn from the sound of it.

Or we could say "well 'consciousness' is already a word with a lot of meaning attached (it's already used in ways which assume certain properties of it) and so we'll continue with that use and if the mental processes we're investigating turn out not to have those properties then calling them 'consciousness' in that sense is incorrect, they seem that way, but aren't". That's the route Dennett takes.

I think his choice of route has more to do with selling books than philosophical merit, but that's not relevant to his conclusions so long as we understand his choice.