Is Integrated Information Theory of Consciousness a pseudoscience?
resolved Nov 1

As claimed by

Resolves by the market. Will not participate, I already have a strong opinion but want to know what people here think and why. Would love it if you leave a comment explaining your bet. Related to this somewhat

Get Ṁ600 play money

🏅 Top traders

#NameTotal profit
Sort by:

perhaps resolution by the maket was (aptly or not) the most pseudoscientific way to resolve this…


I do strongly believe IIT is a pseudoscience. It does ... kind of ... make predictions, but when those predictions weren't compatible with what the believers set out to explain, they just dodge the issue.

predicted YES

Theoretical computer scientist Scott Aaronson has criticized IIT by demonstrating through its own formulation that an inactive series of logic gates, arranged in the correct way, would not only be conscious but be "unboundedly more conscious than humans are."[31] Tononi himself agrees with the assessment and argues that according to IIT, an even simpler arrangement of inactive logic gates, if large enough, would also be conscious. However he further argues that this is a strength of IIT rather than a weakness, because that's exactly the sort of cytoarchitecture followed by large portions of the cerebral cortex,[32][33] specially at the back of the brain,[2] which is the most likely neuroanatomical correlate of consciousness according to some reviews.[34]

predicted NO

@jacksonpolack The market was resolved incorrectly. How much time have you actually spent thinking and reading about consciousness?

predicted YES

This is what 'resolve to market' means (well, that or resolves prob), and this is what happens to every market that 'resolves to market', it was resolved correctly. I agree it's at best an imperfect way to elicit truth. I have thought about this a decent amount and do not think IIT is in any way tenable!

predicted NO


> I have thought about this a decent amount
Can you put a number of it?

predicted YES

IIT itself maybe 2 hours total ever? "Consciousness", easily >50h. It's a big topic.

Also, "IIT is pseudoscience" is mostly the consensus view? The consensus is wrong sometimes but sometimes isn't.

predicted NO


> IIT itself maybe 2 hours total ever? "Consciousness", easily >50h. It's a big topic.
I've spent thinking and reading about consciousness 50h in like last week. IIT itself is probably like 10-15 hours in total.

> Also, "IIT is pseudoscience" is mostly the consensus view?
It's not. If you don't particularly trust Jacob, you can you can check out the QT'd tweet from his thread. There is also this substack post with an in-depth explanation.

Principia Qualia (the best book on concsiousness IMO) says "the current iteration of IIT is probably wrong, and definitely incomplete" which is IMO accurate, but this is far from it being a pseudoscience.

I take an L on this market since it's effectively self-resolving but IIT is not pseudoscience.

predicted YES

Honestly, 'pseudoscience' is just a fancy way of saying 'wrong and stupid', and trying to dissect whether something is pseudoscience or not beyond vibes is not worth it. (edit, tone note: snark aimed at the term 'pseudoscience', not iit)

I don't have a high opinion of most of modern psychology or cognitive science, and agree with jacob there. I just think IIT is also pretty bad. With a few minutes of skimming, I don't think Seth's "Weak IIT" is any better tbh. There seems to be a ton of mistaken ideas that accumulate atop each other and aren't corrected. I think it's analogous to how mistakes accumulate in continental philosophy.

A lot of people in this thread are relying on the Popperian concept of falsifiability to decide if IIT is a pseudooscience, but that is only one way of considering the issue. Without getting into the philosophy of science too much, there are many concepts in science that can be useful as part of the scientific process that are not falsifiable--we call them definitions. You can use definitions as a starting point e.g. to see what properties are associated with other properties. The question of whether a definition is part of a science has to do with whether or not at least some people engaged in science find the definition to be fruitful for asking questions. I think IIT meets this threshold (at the present time). So I do not think it is a pseudoscience.

predicted YES

@Nostr0m Do you take IIT to be a definition of consciousness? Or just a useful concept on its own?

predicted NO

@JosephNoonan In my opinion I do not think high integrated information is sufficient (maybe not even necessary) for a system to have consciousness, but it could still be a useful property of putatively conscious systems to measure (among a set of other properties) that might be associated with other interesting properties. So I disagree with Tononi but still think it could be useful to think about in a scientific context thus I do not think it is a pseudoscience.

I don't consider it pseudoscience. It offers falsifiable predictions about the structure of conscious systems. Like other ToCs. it sometimes shifts goal posts, but it is not formulated in an ungrounded way. Also, the reasoning for it being pseudoscience is shaky: it's just because it predicts things that are intuitively odd (e.g. phi being attributed to all organized matter). But these might either be solvable edge cases or perhaps there is something to it to call these systems conscious to some extent.

I am not an IIT proponent, I can criticize it and prefer other schools of thought, but the qualification pseudoscience should be reserved for bad faith efforts to have ideas fit data and I just don't see that in IIT. I also think the actual explanation for consciousness will lend some ideas from IIT in the end.

@Vincent The question then is, how will this market resolve. This is a question about the minds of Manifolders. I am not sure I want to bet about these :)

@Vincent What falsifiable predictions does it make? If you don't accept, "IIT predicts that this obviously non-conscious system is conscious," then I don't really see how you can say that it has or hasn't been falsified, since the only predictions it makes are about what is and isn't conscious.

I don't think think the term pseudoscience should be reserved only for pseudoscience that's put forward in bad faith. Lots of people genuinely believe in all sorts of pseudoscientific nonsense, and many forms of pseudoscience start off as a result of bad reasoning and wishful thinking, not because someone intentionally tried to mislead others.

@JosephNoonan I agree about pseudoscience, which cannot be defined the way I put it. I think my point there was only that IIT -- which I do not subscribe to -- makes an effort to create falsifiable predictions and therefore falls within the realm of science. I guess the demarcation problem for science vs. pseudoscience needs a resolution for this market, yet I'd offer that IIT's commitment to describing the necessary structure and dynamics of systems that are conscious make it a theory within the boundaries of science.

The problem with "obviously non-conscious" is that this assessment depends on a theory of what consciousness is to begin with, it's just not explicit about it. If the consciousness that we try to understand indeed depends on some concept related to information integration and embedding into an environment, then once we have the definitions sharp we might also see it applies to systems that we had not considered to conscious before. Perhaps plants do have some degree of consciousness - they surely integrate information to ready action plans. I don't think we can serve off IIT just for making such a suggestion, because the conviction that plants would not be conscious is not on firm ground itself, but rather based on a very intuitive and aspecific idea of what consciousness is. Still, that "surely this can't be right" argument is the argument made in the preprint.

As for falsification, if I were to defend IIT I would say that their construct of what is consciousness predicts certain features that can be tested against e.g. neural data. You'd expect a particular form of integration to correspond to conscious states and less of that for less conscious states. Indeed, IIT can describe how sleep and awake states differ in terms of neural activation. I am not too sold on this, so I am reluctant to offer it as some sort of vindication -- the theory was altered to fit existing data -- but in principle the existing work can be tested against novel variations in conscious states. E.g., I recently saw some stuff on IIT measures comparing people engrossed in a movie vs. being aware of themselves/surroundings. For me that's enough to say that there's a real effort to make testable predictions and check whether they are coherent with empirical work (which differs from distinguishing different theories of consciousness), well within the realm of science.

predicted YES

@Vincent But this is exactly the problem. When IIT predicts that systems that are intuitively non-conscious, like plants (or even worse, the example in Adam Treat's link below, which IIT predicts is far more conscious than a human) are actually conscious, proponents say, "Oh, well maybe those systems really are conscious. To say otherwise presupposes a different theory of consciousness." But when an intuitively more conscious state does turn out to have a higher value of integrated information than a less conscious state, they say, "This confirms the predictions of IIT." You could argue for one of the other, but you can't have it both ways. Either IIT's failure to judge intuitively non-conscious systems as non-conscious is a falsification, or its success in describing some intuitively more conscious systems as more conscious isn't a confirmation. Either way, we either find that IIT is already falsified, or unfalsifiable.

@JosephNoonan I do think our intuitions about conscious systems like us (e.g. stating that drowsiness is different from alert in humans) are of a different kind than those about plants. We make it work for systems that are less controversial and then generalise elsewhere. This is not unusual in science, is it?

predicted YES

@Vincent Is it less controversial that we are more conscious when alert than drowsy than it is to say that plants aren't conscious? Some people don't think consciousness comes in degrees at all and so would reject the first claim while still agreeing that plants are obviously not conscious. Even among people who accept the first claim, I think most would say it's less obvious than the second. I would be much more likely to accept a theory of consciousness if it implied that I'm actually just as conscious when drowsy as I am when awake than I would if it said that plants are conscious.

@JosephNoonan Being falsified does not make something pseudoscience. Changing a theory to account for its predictions being shown false does not make it pseudoscience.

You are accusing IIT of nothing which is not also true of superstring theory. Superstring theory is not fantastic science, but it is indisputably science. So is IIT.

predicted YES

@JiSK When did I say either of these things? I have been pretty consistent in saying that the reason IIT is pseudoscience is because it doesn't follow scientific methodology, despite sometimes being presented as science. This has nothing to do with whether it has been falsified, and I don't think I even accused it of adjusting its predictions to avoid being shown false. (Although I don't think you are exactly right about what makes things pseudoscience anyway. Being shown to be false doesn't retroactively make genuine hypotheses pseudoscience, but it does mean that continuing to advocate for them is pseudoscience. And while it's not pseudoscience to make some adjustments to a successful theory, continually adjusting a hypothesis because all its predictions have failed is a hallmark of pseudoscience).

The reason IIT is pseudoscience is because it makes no empirical predictions to begin with. IIT is only a theory of what consciousness is, so it doesn't actually tell us anything about what we should observe in experiments, unless we supplement it with additional hypotheses about what is and isn't conscious. But proponents of IIT are selective about what additional assumptions to supplement it with. They're happy to go with our intuitive assumptions about consciousness when it fits their theory and claim that this is a confirmation of IIT, but when our intuitive assumptions don't fit the theory, they say that this just means our intuitions are wrong. This is not good methodology in any field of study, let alone science.

Superstring theory is not fantastic science, but it is indisputably science. So is IIT.

Did you actually mean to claim that IIT is indisputably science, or was this just a mistake? IIT is indisputably not science. It's widely criticized as unscientific by neuroscientists, and even many neuroscientists who give it some support probably wouldn't go so far as calling it "science". The best argument I've heard against it being pseudoscience is that some proponents don't even pretend that it's science, but treat it as speculative philosophy (and I agree that these proponents are not engaging in pseudoscience). However, I've seen it either misreported as science in the media or treated as science by proponents often enough that I think it can be considered pseudoscience.

predicted NO

@JosephNoonan You said it here, and here

None of your other criticisms are valid either, unless you want to contend that string theory is a pseudoscience as well. IIT is exactly like string theory - an early theory of a subject poorly understood, indisputably scientifically serious, with known problems but none that prevent it from being a fruitful 'current best option' to continue to explore and refine in the hope that it will lead to a more complete theory.

predicted YES

@JiSK None of those were cases of adjusting its predictions after they were proven false. They're cases of using an inconsistent standard for falsification and confirmation. That's what I'm criticizing: IIT proponents claim to have experimental support because certain systems that are intuitively more conscious appear to have higher phi, but when it is falsified by the exact same standard (systems that are intuitively less conscious have higher phi, or systems that are intuitively not conscious at all are conscious according to IIT), they say that it hasn't really been falsified because that's not a good standard. They're trying to have their cake and eat it to.

As far as I know, no string theorist has tried to claim empirical support for string theory using a double standard like this, but if someone has, I would accuse them of pseudoscience as well. And to be clear, I think there is some pseudoscience going on in string theory as well. For example, popular science media sometimes treats string theory as if it's been established, which is definitely a pseudoscientific claim.

IIT is exactly like string theory - an early theory of a subject poorly understood, indisputably scientifically serious, with known problems but none that prevent it from being a fruitful 'current best option' to continue to explore and refine in the hope that it will lead to a more complete theory.

I partially agree with this. Treating it as "the best thing we currently have," while acknowledging that it is not actually true and trying to use it as a stepping stone to a better theory is a good idea. But if it's really going to be considered science, we have to fix the methodological flaws and use consistent standards for experimental tests.

Biting my tongue here a bit, but loving the discussion and would like to add some fuel. Is the problem the IIT part or the consciousness part (e.g. consider the analogy IIT : consciousness :: String Theory : universe ).

I think the word pseudoscience should be reserved for things which pretend to be science despite being demonstrably false; I don't think speculative philosophy counts.

predicted YES

@April I disagree. Pseudoscience is anything that masquerades as science despite not following the methodology of science. Limiting this only to things that are demonstrably false means it's impossible to call things that definitely are pseudoscience pseudoscience. If I've picked up the latest new age healing practice, and no one has yet bothered to perform scientific studies proving that it doesn't work, does that mean I can claim that the mystical healing magic works, and present it as being on par with or better than established medicine, without this being labelled as pseudoscience?

The purpose of the pseudoscience label is to prevent people from passing something off as science when it's really not. We already have a phrase for things that are demonstrably false.

Also, arguably, IIT is demonstrably false (see the post linked by Adam Treat below).

predicted YES

@April But this is exactly something pretending to be science but demonstrably false. Dead logic gates arranged in a particular order are not any more conscious than being born in a certain month determines your personality type.

predicted NO

@AdamTreat Surely if you use enough logic gates you can simulate a human brain

predicted NO

@JosephNoonan Is it the case that ITT masquerades as science? It's always seemed very clearly to be speculative philosophy to me.

predicted YES

@April But the specific arrangement described in the link doesn't appear to be conscious. It's just a rather unremarkable matrix applied to the starting state over and over again.

predicted YES

@April I have seen people portray IIT as if it's science.

@JosephNoonan Hmm. Okay, yeah, I agree that someone who portrays ITT as established science would be engaging in pseudoscience.

I think I disapprove of denouncing the theory rather than misleading presentations of the theory as pseudoscience, though.

"Hey, maybe consciousness works like this? I think it does!" isn't pseudoscience, but "this is how consciousness works" can be.