Consciousness as Representing One’s Mind: Rethinking the Higher-Order Thought Theory of Consciousness

Word on the street is that I am allegedly writing a book on consciousness…woah betide us, this is certainly the final indicator that we are in the most absurd of simulations! At any rate,I don’t have a contract or anything but there is ‘some interest’ in my completing a draft from a press, which is cool I guess (I can think of at least 4 people who would actually read a book like this!). To help motivate me I have decided to put up chapters here as I draft them. My plan is to officially start this summer, since I seem to be only teaching one class in my summer session (I usually do 3 so it feels light). I don’t have anything I want to share yet, but below is the plan of attack i.e. the proposal for the book). Stay tuned!

  1. Brief Description – The cognitive neuroscience of conscious is establishing itself as a viable branch of neuroscience. Currently the field is at the point where there are a variety of theories about the nature of consciousness on offer and we would like to compare the predictions that various theories make to narrow down the plausible candidates. This makes it especially important to develop the candidate theories in enough detail that they could meaningfully confront the tribunal of experience (i.e., be subject to possible empirical falsification). Higher-order theories of consciousness have enjoyed some attention in philosophy and have very recently been taken seriously by neuroscientists aiming to empirically test the theory. However, because of the way that the higher-order approach has been presented a version of the theory -which I think of as one of the most promising versions- has been overlooked. This book will develop and defend this Higher-Order Representation of a Representation (HOROR) theory of phenomenal consciousness. My goal, as stated above, will be to develop the theory in enough detail that it can be empirically compared to other versions of the higher-order approach. Once this is done we can take stock of recent neuroscience and see that the HOROR theory, like a few theories of consciousness, is consistent with current neuroscience. We will also be able to see what kind of neuroscientific evidence we could find that would falsify this kind of theory. Thus whether one is sympathetic to the higher-order approach or not this project will help to clarify what would count as empirical evidence for or against this approach. If one is interested in trying to show that theory is wrong, or in vindicating it, one needs to look at versions of the theory that are taken seriously by people who hold these kinds of theories. 
  2. Outline –       

I. Traditional HOT is mistaken

The science of consciousness is at a point where we would like to narrow the range of theories that are serious contenders. The field has seen a number of theories being presented, from various disciplines, but as of now we have not seen empirical evidence that the various theories can’t interpret in a way favorable to, or at least compatible with, their theory. Theories should be developed in enough detail to see what predictions they make and how we could falsify them. Higher-order theories have enjoyed some support recently but as of yet it is not clear what would count as falsifying the theory. As they are generally presented they (a) aim to explain the difference between a conscious mental state and an unconscious mental state and (b) this explanation takes the form of positing two mental states, one of which is directed at the other. Each of the above claims problematic. The problem with (a) is that it begs the question as to whether there are any unconscious mental states, which is an empirical question and (b) obscures the distinction between relational and non-relational higher-order views. I want to present and defend a version of the non-relation view I call the Higher-Order Representation of a Representation (HOROR) theory. This seems to me to be a promising version of the theory and as of yet has not been developed in detail -because of the way the debate has been set up- and distinguished from other versions of the theory.

II. Starting Over: The HOROR Theory of Phenomenal Consciousness 

Higher-order theories of consciousness appeal to inner awareness as part of the explanation of phenomenal consciousness. ‘Inner’ means higher-order, an awareness of my own mental life (‘first-order’ thus mean awareness of something which isn’t mental). ‘Awareness’ at a commonsense level means sensing, perceiving or thinking that something is present -these are all representational states. ‘Phenomenal consciousness’ means: what it is like to be a creature or what it is like to be in a mental state/ so appealing to inner awareness amounts to appealing to ‘higher-order representations of representations’ -HORORs- as part of the explanation of phenomenal consciousness. I take a theory of consciousness to primarily be a theory of phenomenal consciousness. Panpsychism, Global Workspace Theory, whatever theory one may have, if it is a theory of consciousness, it is a theory of phenomenal consciousness. This makes the theory I will defend not a version of illusionism. HOROR theory is an empirical conjecture about the nature of phenomena consciousness. 

III. Relational v. Non-Relational higher-order theories

Relational theories are familiar from the traditional approach. Non-relational theories deny the traditional account and instead hold that the relevant higher-order representation is itself enough to account for phenomenal consciousness. Non-relational higher-order theories can be understood to be versions of representationalism, but from the higher-order point of view. Representationalism about consciousness, as I will defend it, holds that for any phenomenal experience we have there is some representational content such that having the phenomenal experience consists in having the representational content. Non-relational theories hold that the right kind of content is a higher-order representation with the content ‘I am aware of (perceptible) red in a distinctly visual way’. Having that state is all by itself enough to account for phenomenal consciousness. Higher-order theories can be distinguished by the kind of content they posit at the higher-order level as well as by the proposed relation, if any, between the higher-order representations and their targets, the states they represent. 

IV. Two Kinds of Relational Theory: Joint-Determination and Split-Level

Joint-determination views hold that both the higher-order and lower-order state contribute to phenomenal consciousness. Split-level views hold that the higher-order state is a mere pointer that points to a first-order state, which then contributes to phenomenal consciousness. The difference between these theories is that on Joint-Determination views just having the higher-order state without its target will result in an atypical experience. On Split-level views the first-order state and its content, once it has been pointed to, completely determines what it is like for one but without the higher-order pointer the content remains unconscious and so not experienced. Traditional objections to higher-order thought theories (like those from Dretske on change blindness, the problem of the rock) apply mostly to relational theories. In addition it is not clear that these theories can offer an explanation of consciousness and have to settle for fitting the data. Some find these objections serious enough to disregard these kinds of theories but nature may just foil our desire for understanding and explanation. Relational theories make empirical predications and though I would prefer non-relational theories we should base our credence on the data not philosophical objections.

V. Two kinds of non-relational theory: HOT and HOROR

Traditional HOT (THOT) theory is non-relational. But none the less maintains that the first-order state is the conscious state. HOROR theory holds that the higher-order state is itself the phenomenally conscious state. The higher-order state is itself the state which there is something that there is like to be in. The THOT theory holds that the higher-order state ‘engenders’ phenomenal consciousness but that is difficult to make sense of and the ‘traditional’ objections to higher-order theories can be seen to be problems only for the relational view. For example, ‘the problem of the rock’ which asks why thinking about a rock doesn’t make the rock conscious when it does make my mental state conscious, clearly assumes the relational account of the higher-order theory. Which version of non-relation theory is preferable? HOROR theory is more plausible for several reasons to be discussed.

VI. HOROR theory and Current Neuroscience

There is much debate about the neural correlates of consciousness and there have been some attempts to use empirical work to challenge higher-order theories. One area where this has occurred is with Ned Block’s argument from phenomenological overflow. If consciousness ‘overflows’ what we can report on at any given moment then is that a problem for higher-order theories? Do we have any reason to believe that there is phenomenological overflow in this sense? HOROR theory is compatible with either view in this area. We could have a ‘Rich’ HOROR theory on which the rich contents of the higher-order states overflow what we can report or what is in working memory. We could also have a ‘sparse’ HOROR theory on which there are sparse contents of HORORs. Because of this there are versions of HOROR theory that are compatible with either way you interpret the findings from Sperling, Landman, etc. Part of the problem is that we haven’t really seen scientists explicitly try to falsify versions of non-relation theories and so we need to get clear on what kind of predictions the theory actually makes. The same issue arises with other previous empirical attacks on the higher-order approach (like Dretske’s change blindness argument, and Ned Block’s response to Lau and Brown’s use of Rahnev et al as evidence for HOROR).

VII. Empirical Predictions: A Study in Misrepresentation

Misrepresentation occurs when HO state misrepresents a first-order state. Radical misrepresentation occurs when FO state missing. Misrepresentation can be seen as an empirical predication of higher-order theories. Split-level views predict there will be a first-order explanation of misrepresentation. Joint-determination views predict there will be a ‘partial’ experience. Non-relational views predict conscious experience follows higher-order representation (and could be sparse or rich depending on the contents of the HORORs).THOT has to say non-existent first-order state is conscious! One can make sense of this with certain views about intentionality and representation but those views are very controversial. HOROR theory says misrepresentation is evidence that it is the higher-order representation which is phenomenally conscious. Radical misrepresentation is a case which shows that the state which there is something that it is like for one to be in is the HOROR itself. The overall lesson from thinking though cases of misrepresentation is that we should postulate that there are two kinds of content in the relevant HORORs: a descriptive content and a pointer/teleological/causal-historical content. The descriptive content accounts for phenomenology while the other content accounts for which first-order state is picked out. This pointer kind of content will have a functional role (keeping a represented state online, sending it to the global workspace, etc). This puts HOROR theory in between relational views on one end with solely pointer content and THOT on the other with solely descriptive content. Thus different versions of relational and non-relational higher-order theories can be tested by looking at/for cases of misrepresentation.

VIII. Implementing the Theory in the Brain

The theories as so far presented is a psychological theory. By itself they make no predications about how these various kinds of states are implemented in the brain. We have various proposals about how the implementation should go. Starting at the first-order level we might ask, ‘where are the first-order states, the targets?’ We have several candidates: recurrent processing, mid-level contents, globally broadcast contents, etc. Some of these candidates are in ‘sensory’ areas, others may be in frontal cortex (there is evidence that representations we would think of a first-order are in the prefrontal cortex at least sometimes). So it is not yet clear where the first-order states are. What about higher-order states? Lau argues they have overlap with metacognition and so prefrontal cortex. LeDoux argues anatomy suggests different circuits in prefrontal cortex with different jobs. Genaro suggests that we look for self-consciousness and focus more on parietal cortex. Cleermans argues anywhere in cortex. Not clear where higher-order states are. This doesn’t mean we can try to empirically test these theories.  We should formulate different versions  in as much detail as we can and test them (like the Prefrontally-implemented Rich version of HOROR theory Templeton is testing). To really test these theories though we need a hypothetical ‘Brain clamp’ -something that allows us to hold the activity of the first-order representations constant while we vary higher-order content (and vice-versa).

IX. Animals, Infants, and Robots

The discussion so far has centered for the most part on adult human beings and sought to develop a possible account of the kind of conscious experience we enjoy. But can these ideas help us answer questions about whether animals are conscious? Can we know if infants and newborns are conscious? Is artificial consciousness possible? And what does the HOROR theory predict about these kinds of cases? Can we use animal models to test the HOROR theory? These questions may be somewhat more speculative and less connected directly to the issue of empirical testing but the higher-order approach has been thought to have a certain position on these questions. My own view would be that animals and infants are conscious and *maybe* we could have artificial consciousness but that as of right now we don’t have any real strong evidence that this is the case. However the ideas presented here would at least give us some ideas about how we might -at least in principle- be able to empirically test HORORs predictions in animals and infants. 

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s