I have just come into contact with the integrated information theory of consciousness (and its relatives), which would have it that consciousness stems from information-rich (and integrated) systems.

My intuition tells me that two different systems carrying the same complexity of information should be equivalent under these theories, and (equivalently, I think) that one cannot discriminate between two conscious experiences that contain the same amount/complexity of information (i. e. one cannot say that such a system is conscious of a picture of a flower rather than of a verbal thought under the assumption that these require the same amount/complexity of information).

My "reasoning" is that a string of information which may be interpreted as encoding a picture may under a proper choice of language instead be interpreted as encoding a novel. Would you agree?

  • This isn't my area of expertise, so I might be missing something completely. But just because two things have equivalent complexity does not mean that they are equally apt for capturing the same information. Or to put it another way, if you ears could hear the same # of frequencies as your eyes can see, that doesn't mean both would capture the same information. It just means that which they would capture would be equally complex.
    – virmaior
    Jul 16, 2014 at 1:42
  • "Pure" complexity is not the model that IIT is built on. Its author G. Tononi explicitly separates between different neural arrangements which may appear to have equal complexity but have distinctly different connection patterns, which he presents as a reason to distinguish between cortical areas of the brain and the cerebellum. Moreover, the paper at biobull.org is a few years behind Tononi's latest revision of the formula for ɸ. If you've just stumbled on IIT, I highly recommend catching up here: scottaaronson.com/blog/?p=1799 The comments and follow-up posts are invaluable.
    – Ryder
    Jul 16, 2014 at 15:16

2 Answers 2


Your theory suffers from some serious ambiguities. The first problem is that you haven't explained what is the relevant measure of complexity and why. This is a more difficult issue than it might appear to somebody who hasn't given it much thought. Charles Bennett has discussed these issues in many papers, such as


For example, suppose that you decide to describe the amount of information in an object by the compressibility of a description of an object. The first problem is deciding the relevant level of description, to which I will return later because it reveals deeper problems with your position. The second problem is that using standard compression algorithms the most complex sequences are random, e.g. - a series of coin tosses.

The best measure of complexity I know of that is relevant to knowledge bearing systems is Bennett's own suggestion: logical depth. The complexity is the number of steps that a universal computer would require to produce the relevant description. This avoids the coin toss problem if you include a random number generator in the computer. Random number generators are allowed by the laws of physics, e.g. - measurement results of whether a particular atom has an electron in a particular shell, so this seems reasonable.

Now, you say that any description can be interpreted as representing any other description. There is a sense in which this is true, but it would destroy your whole position if you took it seriously. For if you allow an arbitrary mapping between the description and the things to which it refers, then a sequence of random coin tosses instantiates a novel exactly as well as, say, the contents of your genes or the complete works of Shakespeare. What is required for any measure of complexity that makes sense is that you include the information required to do the relevant interpretation in the description and then consider what the item in question adds in terms of complexity. So for Shakespeare you might have to include knowledge of grammar, aesthetics, history and probably other stuff. For a gene, you would have to specify some features of the environment, including what other genes are present. For a random series of coin tosses you have to add how to identify the different sides of the coin.

To describe conscious experiences, I don't think anybody knows what the relevant description would be, nor do I think it would be the same for everybody since different people have different dispositions, ideas and so on. For an example, see


For some idea of many of the problems involved in coming up with such a description see "Consciousness Explained" by Daniel C. Dennett.

The idea raised in your question doesn't make sense. But there are many deep problems associated with the problem of how to specify complexity that are worth investigating.

  • Thank you for your answer! I have indeed little knowledge of the subtleties of information complexity, and I'm still a bit confused: My question was intended to express doubts about the position that consciousness is all a matter of information complexity, rather than defending it. At the heart is your observation that any large random string may be interpreted as containing a novel, but to actually interpret it would require additional information. However, if pure information gave rise to consciousness intrinsically, it should do so independently of the interpretation. Still makes no sense?
    – DanielF
    Jul 16, 2014 at 11:49
  • Pure information doesn't give rise to consciousness because pure information doesn't exist. Any piece of information has to have some interpretation in the sense of a mapping between the information and whatever it refers to. For example, the string "(=<#9]~6ZY32Vx/4Rs+0No-&Jk)"Fh}|Bcy?=*z]Kw%oG4UUS0/@-ejc(:'8dc" seems not to mean anything, but in fact it is a program for printing "Hello World!" in the programming language Malbolge, see en.wikipedia.org/wiki/Malbolge#.22Hello_World.21.22_in_Malbolge
    – alanf
    Jul 16, 2014 at 12:51
  • OK, to my limited understanding the integrated information theory claims that "pure" information defined via some kind of entropy and independent of interpretation is what gives rise to cinsciousness. Have I misunderstood the claim of the theory, or is your comment intended to be a refutation of it? Thanks again.
    – DanielF
    Jul 16, 2014 at 13:08
  • Having looked quickly at the paper describing that theory biolbull.org/content/215/3/216.long, it doesn't say anything about pure information. Some remarks in the paper contradict the pure information position, see paragraph 2 of "Consciousness as an intrinsic property" . It does say that consciousness can be quantified by the entropy of some information in your brain, which is false since entropy is not a good measure of complexity. It assigns a large amount of information to a random sequence. So the theory is false, but not because of the idea of pure information.
    – alanf
    Jul 16, 2014 at 13:40

OK, so I think I have found a discussion (between Searle and Koch/Tononi) concerning precisely the point I wanted to raise: http://www.nybooks.com/articles/archives/2013/mar/07/can-photodiode-be-conscious/ . Koch/Tononi basically claim that the fact that their information is supposed to be "integrated" (and "non-Shannonian") renders this criticism invalid, while Searle disagrees.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .