finale: reply to Gomez-Marin & Seth about the scientific status of IIT

here i respond to a commentary written by Alex Gomez-Marin and Anil Seth, on a paper explaining why IIT is unscientific. the latter article was co-authored by 100+ authors including myself.

***

more than just an issue of testability: Gomez-Marin and Seth wrote that our case against IIT primarily concerns testability, and suggested that ideas not currently testable may become so in the future. but we already explicitly acknowledged this point in the paper (“Making theoretical predictions that may not be testable by current methodology does not necessarily render the relevant theories unscientific. However…”). importantly our concerns are also definitely not only about current testability. other problems include our worry that some of the core claims that set the theory apart from others can never be testable, becoz they make no observable difference even in principle. 

further, although the foundation of the theory is presented as axiomatic, it involves many arbitrary components. there has also been a considerable degree of metaphysical overreach (e.g. claims concerning free will, and that neurons don't really exist). as a research programme, it also seems to have been rather resistant to falsification (their NCC predictions changed over time, apparently sometimes after pre-registration and data collection, and they also seem unmoved by the exposition that according to their theory some sets of inactive logic-gates are conscious). 

overpromotion: this is also a significant concern. Gomez-Marin and Seth made the point that overpromotion should not change a theory’s scientific status, and that “science isn’t a popularity contest”. but i find the rhetorics confusing. on this point i thought we have been pretty clear: “Although misrepresentation itself does not make the theory unscientific, it suggests that the label ‘pseudoscience’ may be appropriate.” as we have provided independent arguments as to why the theory is unscientific, this is exactly in line with the definition that Gomez-Marin and Seth themselves provided: “pseudoscience is work that purports to be scientific, but falls short in some substantive way or ways.” (emphasis mine)

Lakatos & secondary predictions: one point that may appear to have more substance concerns Lakatos’s account of a non-degenerate science: that "it generates testable predictions with explanatory power, which in turn generate more such predictions, in a productive iteration" (emphasis mine). Gomez-Marin and Seth claim that IIT generates such predictions. but i do not find this claim convincing. 

their first example is a controversial study in which IIT predicts that we can decode visual information from the visual cortex, under normal viewing conditions (i.e. no control for nonconscious perception). as the result is rather trivial, i’m not sure what additional explanatory power the theory provides. the study also suffered from several other problems. at the time when the project commenced, Seth himself commented that it would not "do what it says on the tin”.

the other example is an ongoing study that is funded through a similar mechanism, which i consider problematic. here IIT predicts currently non-firing neurons could make a difference to perception, if they are inactivated. but i’m not sure this is such a novel and counterintuitive prediction as the authors claimed. rather, it is a common understanding nowadays in neuroscience that population coding is the norm. for a whole population of relevant neurons to contribute, of course many will be silent during a perceptual event. that this could have perceptual consequences is readily explained by any well-functioning downstream readout mechanisms that are sensitive to not just active inputs, but also the lack of inputs from some upstream neurons. if these ‘silent’ neurons are inactivated, so that they don’t just not fire on specific occasion, but just couldn’t fire anymore within a broader context, any reasonable system may well adjust accordingly. 

i.e. we see red as red becoz the green-coding neurons for the same stimulus are not signaling just as positively at the same time. if a sensible downstream readout mechanism cannot ascertain whether the green-coding neurons are actually voting ‘no’, or just couldn’t vote at all (e.g. because they are shutdown, or dead), why should we be surprised if it makes a difference?

therefore, i’m not sure these examples of relatively banal ‘predictions’ are in line with Lakatos’s account of productive/progressive science at all.

***

a right to wrong? - a point made in the article has previously also been made by Seth, in his plea for the “right to be wrong”, to which i have already responded independently. of course, in science we must be allowed to be wrong sometimes. but this is also where the point of overpromotion matters. being wrong is one thing, telling the public that it is the most promising, only game in town, while consuming tens of millions of dollars of research funding is a different matter. meanwhile, if you’re widely perceived to be wrong or not even wrong, you just can’t in good conscience tell trainees that your theory is widely recognized to be the exact opposite. that would be asking for a right to wrong, rather than a right to be wrong, as trainees deserve to know what they are getting into, rather than be misled into potentially wasting their time and career on what is widely considered to be deeply problematic. they deserve to know about the different opinions out there, in order to be able to make up their own minds. our article is just to set this public record straight.

humility & what caused the reactions in the first place: from there, Gomez-Marin and Seth also mentioned that for understanding consciousness, humility is a relevant virtue. to this i wholeheartedly agree. if IIT had been correctly (and humbly) presented as a speculative idea that is far from well established (and possibly much worse), there would have been no need for us to write our article. and yet, the hyperbolic promotion has been intense, and frankly, at times just over the top. everybody within the field, as well as many outside, are fully aware of how far things have gone. it is puzzling that some researchers seem to be unwilling to acknowledge what really rendered our protest necessary in the first place. even if they have themselves also contributed to this problematic promotion, for the benefit of new trainees who do not have access to insider gossip, the need for public correction should take priority over the desire to rationalize and to justify their own past actions. 

“trial by authority”? -  Gomez-Marin and Seth also brought up the example of astrology, and cited why Carl Sagan refused to sign a letter against it, despite his belief that it lacks scientific basis. for me, if IIT had already been publicly characterized in the same way that astrology has been, i would also not sign statements that criticize it further. it is also ironic that researchers arguing against the nature of these co-signed statements of concern often themselves sign similar statements when they see fit. besides this letter which Seth signed, there is also this ‘declaration’ on animal consciousness, which he also signed, together with numerous proponents of IIT and those who objected to the very nature of our letter. to my mind, the declaration (including the supporting background document) lacks both relevant substance and sound logic, and will likely hinder scientific progress for far beyond consciousness research alone. (notice that the letter and the declaration, both of which signed by Seth, also contradict each other on the status of animals without cerebral cortex, at least in terms of the spirit of the arguments.)

changes in number of authors: finally, Gomez-Marin and Seth noticed that the number of authors seem to have gone down from 124 to 102. this is because a mere signature in support of a statement of concern is different from co-authorship for a published journal article (which is especially time consuming given the number of authors involved). this is explained in Q3 of this FAQ document, which also provides further background details for the initial letter / statement of concern as well as the latest article.

***

i probably won’t be commenting on IIT in writing anymore from here. to my mind, it was ever only a symptom. the field has deeper systemic issues, and i explain my broader concerns here. while i’m still interested in mechanistic explanations of the differences between perceptual processing that come with and without subjective experiences, i no longer think that participating in the current research community would help us achieve that.

so this will also be my last post on this blog. 

perhaps we can’t really trust ‘consciousness’ after all.

on being taken seriously

yesterday i posted this on twitter. didn't think of doing it here coz it's a rather simple point. basically i think some panpsychists have been making claims that i find less than intellectually honest, and i have made a complaint to SEP like i said i would. SEP responded promptly and made the author(s) change the text, so that's that.

the claim concerns whether panpsychism is taken seriously by neuroscientists these days. to my mind the obvious answer is no. but there is a tricky sense in which if you can find a couple of neuroscientists who support the view, the positive claim is logically satisfied

this was exactly what Dave Chalmers said on FB, when i brought up the issue in public first. yesterday i referred to the exchange as 'low rhetorics'. but i was promptly told that Dave was actually also part of the editorial process that lead to the change of the text. so he probably didn't intend for it to be a very serious or strong defense. i made a note of clarification and apologies on twitter then.

that said, i didn't know it back then. the point was something said in public. and others e.g. Neil Levy joined in to defend it. so i went ahead and shot SEP the email. just as expected, there wasn't much argument about the case.

so there's little to be surly about. but there's a small part of the argument that is perhaps interesting. so Dave's point was if X is taken seriously by a few members of Y, it is *logically* fine to say X is taken seriously by Ys. but many things are logical to say, yet silly, e.g. 'if 2+2=9, then I am a better philosopher than Dave Chalmers'. certainly for a place like SEP, we'd want the content to be not just logical but also non-silly.

in particular i suggested that it may make sense for one to at least restrict the statement to cases when X is taken seriously by 50% of members of Y or more. Dave and others suggested maybe that's too high a cutoff. perhaps 20% would pass the mark for non-silliness (or something like that).

but i'm not sure. let's say 20% of members of Y take X seriously. so we say 'X is taken seriously by Y'. but it also means 80% of Ys do NOT take X seriously. so certainly, it is just as logical to say 'X is not taken seriously by Y'. so we should allow people to say 'X is taken seriously by Y, and X is also not taken seriously by Y'. or we can say that 'X is taken seriously by Y; the negation of this statement is also true'. or: 'it is both true and untrue that X is taken seriously by Y'.

that's just ..... silly.

but anyway, silliness aside, i don't really feel so strongly about it; mostly just brought up the above for fun. in large part becoz i think it is already a lost cause. some authors will find other rhetorics to promote the view, as they already do. an easy way would be so say the view is endorsed by some 'leading neuroscientist'. again, leading is a pretty subjective thing. there is a sense in which e.g. Christof Koch is a 'leading' neuroscientist, just as Christof likes to unilaterally profess that his preferred theory is 'dominant', 'leading', etc. there's not much to do about it other than to shrug. i too think those old Crick and Koch papers are important and positively influential. but i am not sure many of us think that Christof these days still represents the field we're in. so the situation may be a bit like citing Eccles to say that some of the most accomplished neuroscientists (Nobel laureate no less) were dualists.

yes, yes, you can say that. there are only so many complaint emails i would care to write, and not all editors are reasonable. but why would you want to do that? if you have to cheat this much to promote your view it probably just isn't worth promoting or having, is it?

ok enough about sociology. i may eventually write a post about other problems of panpsychism too. most of us don't like it for scientific reasons, but over the past months i did dig into the philosophy a bit. but i must say, there i'm not so impressed either.....

ps - someone suggested that there may be a difference in standards between science and philo re: what count as being taken seriously. i think that's right, philosophers are meant to consider far-fetched ideas more carefully; that's part of their job. i ran some twitter polls which seem to confirm that too. but if that's so, they should still be careful in saying that certain *scientific* discipline takes their ideas seriously. that's not how it works for us. if we feel misrepresented, it will just get harder for us scientists to take them seriously. and this may be hurting not just panpsychism but philosophy the discipline as a whole. 

pps - this version is updated on 2024 May 9. a previous version implied that one of the authors Philip Goff didn't respond to me promotly enough, which led to my writing directly to the editors. but how it went might have more to do with my own impatience. at some point he said he needed to consult with Dave Chalmers, but i felt that the main SEP editors are somewhat responsible too so writing to them directly was hopefully not inappropriate.

zombies

Zombies are hypothetical creatures which are functionally identical to us, but they lack qualia altogether. If they are possible, functionalism can’t be right. If such possibilities are conceivable, perhaps it means that functionalism is at least conceivably wrong.

There are a few senses in which zombies can be functionally identical to us. Suppose we merely mean they can refer to and act on the same objects in the world like we do. Then sure some robot-like creatures can interact with objects in the world without being conscious.

But if we mean that they are functionally fully identical to us, as in the whole mental algorithm is consistent with what we have, then we should recognize that they can’t be coherently conceived to be nonconscious. Such ‘zombies’ will know what it is like to have certain experiences. They can tell you scarlet is a bit like crimson, not much like pink or purple, etc, from a subjective point of view. This means, if you ask them if there is something it is like to see colors, they will truthfully say yes, for the same reasons we do. On the other hand, if you ask them what it is like to experience some subthreshold firing of neurons in the visual cortex, it should draw a blank. And this would not be merely because they don’t know enough about neuroscience. Even if you actually stimulate their visual cortices that way, they still wouldn’t know what you were talking about. In this sense, they are just like us.

Of course, the anti-functionalist can say that they only behave or talk as if they consciously see. But how else do you know if anyone is conscious, besides their expressing that they think so and your lack of reason to doubt that they are lying or deluded? 

Let’s say we have some Martians, who truthfully agree with us that there is something it is like to see red and other colors. And yet, when asked what it is like to smell, it draws a blank. If odors could nevertheless influence their behavior, we probably would think that they have olfactory processing. In that case, what else apart from consciousness can account for the difference between their vision and olfaction? According to them, in one modality, there is something it is like, and in the other, there isn’t. If they aren’t lying or crazy, what else is the difference?

This is to say, if a creature is capable of thinking meaningfully whether there is something it is like to have a certain experience, then whether the creature is inclined to truthfully think so is the truth of the matter. As such, zombies in a fully functional sense are simply not coherently conceivable. Either they truthfully think there is something it is like to have some experiences, in which case they aren’t zombies; or they do not think so, in which case they are not functionally like us at all.

[this literature is vast and mine will never be the last words, of coz. hence this is a blogpost not a paper. my goal is just to hopefully convince some of my scientist colleagues that some functionalist defense is possible, and it needn't be overly complicated.]