A Few Problems with the “Magic” of Emergence and Supervenience

It’s been argued that the concepts of supervenience and emergence lack “explanatory power”. They’ve even been classed as “magic” (e.g., by Mark A. Bedau). It’s certainly the case that they’re problematic. However, perhaps they’re no more problematic than many of the other options in the philosophy of mind. This piece focuses on some problems, as particularly seen through the philosophy-of-mind positions adopted by the philosophers John Searle and Jaegwon Kim.

[All the quotes in the following are taken from John Searle’s The Rediscovery of the Mind.]

Those who believe in emergence, whether of the mind/consciousness or anything else, don’t escape from the physical. Hence the actual terms “emergence” and “supervenience”: i.e., emerge from the physical, and supervene on the physical.

Although emergence and supervenience may be viewed as “magic”, the emergent phenomenon is still regarded as being utterly dependent on the physical. Mind-body dualism, on the other hand, has it that the substance that is mind is independent of the physical. That means that one of its biggest problems is that it must account for the interaction between the physical and the non-physical. That said, the emergentist must also account for the emergent property even if it is utterly dependent on the physical. In other words, simply arguing that the emergent phenomenon is “dependent on the physical” doesn’t really tell us what kind of relation the one has to the other. After all, the emergent phenomenon usually bears little relation to the physical elements that give rise to it.

John Searle

In detail. The American philosopher John Searle gives a technical account of supervenience in the following passage:

“Intuitively, what is meant by the claim is that mental states are totally dependent on corresponding neurophysiological states in the sense that a difference in mental states would necessarily involve a corresponding difference in neurophysiological states.”

So even if this exact parallelism (although this isn’t parallelism in its historical garb — see here) shows us that mental states are “totally dependent on corresponding neurophysiological states”, and changes in the physical bring about changes in the mental, that still doesn’t really explain the emergence of mental states or the supervenience of mental states. So, although it isn’t presented as a kind of dualism, perhaps it is.

In very simple terms, then, how do we get from a neurophysiological state to a mental state?

John Searle on Causal Powers

One answer to the questions above is to argue that it’s wrong to think in terms of physical states causing mental states.

For one, it may be assumed that if x causes y, then y must have (to use John Searle’s term) “causal powers” too. y doesn’t need to display causal powers immediately after being caused by x. However, surely it must have causal powers (as it’s often put) “in principle”. In other words, isn’t it difficult to imagine a causal x having a relation to a non-causal y? More relevantly, a causal brain having a relation to a non-causal mental state seems intuitively odd. And, as is well known, the causal nature of mental states is much discussed and controversial.

Jaegwon Kim

That is why the philosopher Jaegwon Kim asks questions about epiphenomenalism [see here].

Firstly, Searle puts Kim’s position in the following way:

“[ ] Kim claims that we should not think of the relation of neural events to their supervening mental events as causal, and indeed he claims that supervening mental events have no causal status apart from their supervenience on neurophysiological events that have ‘a more direct causal role’.”

An immediate reaction one may have to that passage is to state that if “the relation of neural events to their supervening mental events” isn’t causal, then perhaps we actually have an identity instead. In other words, if an epiphenomenal y has no (independent) causal role to play, then perhaps that’s because it is x. Now saying that “y is x” still doesn’t tell us why x and y are deemed to be different. More basically, it’s hard to know what the causal relation between x and y is. More relevantly, why is the word “cause” used at all in this particular case?

Searle himself doesn’t have a problem with causation in this respect. He writes:

“It seems to me obvious from everything we know about the brain that macro mental phenomena are all caused by lower-level micro phenomena.”

It’s not obvious what the words “It seems to me obvious” are doing here. Isn’t it the case that even if we knew everything about the brain, any causal relation between the brain and “macro mental phenomenon” still wouldn’t be obvious? What is obvious, however, is that there’s a relation of some kind between the brain and mental phenomena. Indeed, those much-mentioned “mere correlations” between mental states and brain states offer us that. However, is it correct to view these correlations in terms of causation? That said, it’s difficult to think of any alternative to causation in these respects. After all, if the relation between brain states and mental states isn’t a causal one, then what, exactly, is it? So perhaps not even epiphenomenalism escapes from this problem. Moreover, an epiphenomenalist accepts that brain states cause mental states and still argues that mental states are incapable of causing anything.

According to Searle’s take on Kim, the latter argues that “the relation of neural events to their supervening mental events” isn’t causal. Thus, it’s not a case of brain states being causal, and mental states not being causal. It’s a case of not thinking in terms of causality at all. Again, does lead to some kind of identity theory? Alternatively, and as already stated, simply deeming mental states to be supervenient upon brain states doesn’t give us much. What is a supervenient mental state? What is its role and nature?

Downward Causation?

Just to be clear. Searle himself does believe that “mental features” have a causal nature. (An example of downward causation?) He writes:

“[T]he fact that the mental features are supervenient on neuronal features in no way diminishes their causal efficacy. The solidity of the piston is causally supervenient on its molecular structure, but this does not make solidity epiphenomenal; and similarly, the causal supervenience of my present back pain on micro events in my brain does not make the pain epiphenomenal.”

Isn’t the comparison between mental features and the solidity of a piston, and neuronal features and the molecular structure of a piston, just inaccurate? For a start, the molecular structure of a piston constitutes its solidity. Do neuronal features constitute the mental features? In other words, there’s a literal identity between the molecular structure of the piston and the solidity of the piston. Is there a literal identity between neuronal features and mental features? Perhaps there is. However, wouldn’t this make Searle an identity theorist of some kind? Yet he can’t be an identity theorist because he stresses the “causal efficacy” of mental features in distinction to the causal efficacy of neuronal features. An identity theorist, on the other hand, couldn’t argue that “mental features are causally efficacious”, and also that “neuronal features are causally efficacious”, because that would almost amount to a tautology because they’re one and the same thing.

One way around this would be to adopt a position of conceptual dualism and ontological monism (à la Donald Davidson). This means that a given x can be seen under two very different concepts, with the two concepts referring to the same thing. But adopting this position would commit Searle to also believing that mental features have no emergent properties — even though they do fall under concepts which neuronal features don’t fall under. Isn’t it the case, then, that adopting a position of conceptual dualism alongside ontological monism rules out any physical or ontological (rather than conceptual) emergence? In other words, there may well be conceptual “emergence” if that word is used loosely. However, there is still neither physical nor mental emergence.

Is Wittgenstein’s Language Games Idea Both Conservative and Relativist?

The word “conservative” is used because if a language game (to quote Ernest Gellner) “constitutes [its] own automatic vindication”, then surely there’s little room to criticise the status quo. In terms of the word “relativism”. Again, if our language game is the “foundation of the legitimacy of our ideas, in morals, science, politics, anywhere”, then that must apply to all other languages games too. What’s more, isn’t this stance a “kind of romantic communalism, a view that the spirit of the community (expressed in its language) is the only foundation one can ever have for the basic principles of one’s activities, whether moral, aesthetic, scientific or any other”?

Wittgenstein’s ideas about language games belong to his “late period”. However, both periods stress the importance of language. In terms of the following essay, Ernest Gellner says that “[i]t was as if [Wittgenstein] knew how to do two things only, either observe linguistic custom, or project logic onto the world — and having got tired of the former, there was nothing else to do but return to the latter”.

There’s been quite a debate on whether or not Ludwig Wittgenstein’s concept of “language games” is relativist in nature. Oddly enough, some commentators simply ignore the possible relativist implications of Wittgenstein’s philosophy. It’s as if it would besmirch his name to even bring up the possibility of relativism. The philosopher and social anthropologist Ernest Gellner (1925–1995) picked up on this when he wrote about what he calls “Wittgenstein and his intellectual progeny”. (As found in his paper ‘Three Contemporary Styles of Philosophy’.) In technical terms, one way in which our philosophical problems are “dissolved” (a word that has often been used) is with

“the idea that our verbal custom could be the foundation of the legitimacy of ours ideas, in morals, science, politics, anywhere”.

That position, at least on the surface, certainly has a relativist timbre to it. So it’s not surprising that Gellner saw “our custom [as a] problem and not a solution”. He didn’t mean his or our particular custom. He meant any custom at any time is bound to create philosophical and otherwise problems. (It’s worth noting that Gellner never actually mentioned relativism.)

Gellner cited “morals” in the quote above, and later in his essay he went into more detail. Interestingly, he tied his discussion about language games to John Rawls’s thought experiment about the “veil of ignorance” (which is, in my view, an unexpected link). Gellner wrote:

“Are [the values] chosen in this imaginary situation because they are good values, or are they good values because, by definition, anything chosen in this situation is good? Do you *know* that other people, who are unlike yourself, would also choose them?”

It does seem very odd — or at least very radical — to argue that “verbal custom[s] [are] the foundation of the legitimacy of our ideas, in morals, science, politics, anywhere”. Thus, the Wittgensteinian argument must be that whenever a supposed “problem” does arise, then we must consult verbal customs for an answer. Not only that, once we do that, then we’ll quickly realise that it’s not, in fact, a problem at all.

As already stated, this is a very radical — or extreme — position. It also has clear relativist implications. However, many Wittgensteinians have strongly denied this. Gellner puts the almost-obvious case against this denialism when he wrote the following:

“[T]he morality of any given age may indeed be embodied in its verbal behaviour, but obviously this doesn’t mean that we must uncritically accept the morality of any given age or culture.”

At first sight at least, the Wittgensteinian position seems wrong. So perhaps those Wittgensteinians who deny that Wittgenstein was a relativist are right. Yet many people have indeed held this position or similar ones, whether postmodernists, social constructionists, or whoever. That said, the Wittgensteinians who don’t deny this may now ask the following question:

If the “morality of any given age” is not to be found “in its verbal behaviour” (or “verbal custom”), then where, exactly, is it to be found?

Well, one need to do a lot of philosophising to offer an alternative to this. And it can be argued that Wittgenstein himself didn’t like this kind of philosophising.

Gellner offered his readers a reason why (some?) Wittgensteinians have accepted this relativist position. He argued that such people

“did expressly embrace a kind of romantic communalism, a view that the spirit of the community (expressed in its language) is the only foundation one can ever have for the basic principles of one’s activities, whether moral, aesthetic, scientific or any other”.

Gellner quotes the philosopher Stanley Cavell putting this position in even stronger and simpler terms. In terms of Wittgenstein’s “forms of life” at least, Cavell told his readers that “[h]uman speech and activity, sanity and community, rests upon nothing more, but nothing less, than this”.

Gellner’s words (i.e., above the Cavell quote) come across as if he’s expressing a Heideggerian position — or even a National Socialist (Nazi) one. This is made even clearer in a later passage in which Gellner tells us that language-games theory “has a most exaggerated sense of and respect for cultural systems, their autonomy and incommensurability”. Indeed, such Wittgensteinians “endorse[] and love[] them all”. It also “assumes that [language games] are somehow self-contained and authoritative”. Of course, when you scratch the surface, one often finds that even hardcore self-described relativists (the few that there are) don’t accept at least some “forms of life” or “cultural systems”. (For example, the Nazi form of life, the racist form of life, etc.)

Is the Wittgensteinian point here that philosophers, or anyone else, can never transcend (or rise above) “the spirit of the community (expressed in its language)” in order to formulate (or discover) an alternative position on moral, aesthetic, scientific, etc. problems or issues? After all, such people will be doing their transcending in the language — and even in the “spirit” — of the “community” they’re questioning. That said, there’s nothing inherently contradictory about criticising a language with that language, or criticising a community from within that community.

Gellner puts this language-games position in basic terms when he says that if one accepts it, then one must also accept that “cultures are self-legitimating, and validate the norms of conduct and sanity found within them”. Put in another way, Gellner told us that each language game “interacts with the world only in part referentially, but in the main socially, and each of which constitutes its own automatic vindication”.

Thus, surely languages games would include head-hunting cultures, Nazi culture, etc., just as much as they would include the cultures of Sweden and France in 2025, or the United States in the 1960s. (One can ask here is there’s even such a thing as — in the singular — the culture or the language game of any given country at any given time.)

Where is the World?

In terms of Wittgenstein himself, Gellner tackled Wittgenstein’s (possible?) relativism within the context of him rebelling against his Tractatus position. In simple terms, Gellner sees both the Tractatus position and the language-games position as being equally extreme. Talking about Wittgenstein himself, Geller says that

“[i]t was as if he knew how to do two things only, either observe linguistic custom, or project logic onto the world — and having got tired of the former, there was nothing else to do but return to the latter”.

In response to those words one can even say that, at one point, Wittgenstein — among many other philosophers — wanted “the world” to tell us what to say about it. What’s more, philosophers and everyone else must simply be faithful to “the world’s own logic”. This project is suspect in that, even if philosophers were obedient to the world (or at least tried to be), what was to stop them getting it wrong? Perhaps late Wittgenstein realised that philosophers couldn’t really get the world right or wrong because the world itself isn’t made up of statements and theories which can simply be repeated (or communicated) by diligent philosophers. Thus, the ball was always in the “linguistic custom” corner.

So Where is the world within language-game theory?

Gellner didn’t actually believe that (to quote Richard Rorty) “the world is well lost”. At least not when it comes to, of all technical terms, “reference”. He told us that when it comes to language games, “each of which [ ] interacts with the world only in part referentially”. What of the other parts? Language games “interact with the world [ ] in the main socially, and each of which constitutes its own automatic vindication”. Readers may now wonder why a language game needs to be referential at all. (We’d need to flesh out the notion of reference to answer that question.) Can’t a Wittgensteinian language game be built on fictions, untruths, fantasies or hallucinations? For example, why wouldn’t, say, a Harry Potter language game also “constitute[] its own automatic vindication”?

Reference is said to (quoting Keith Donnellan) “tie us to the world”. So, is (to use Gellner’s example) “the nature of things” the source of that tie? Yet, according to Gellner’s Wittgenstein, the nature of things is also an “artefact[] of customs of linguistic communities”. Moreover, this position is directly linked to another of Wittgenstein’s ideas — “meaning is use”. Thus, “[i]f you want to know what a given expression means, don’t ask what it refers to, ask what it does”. In addition, one must “[e]xplore the social context in which [the expression] operates and the multiform purposes which it serves”.

Earlier, Gellner was quoted arguing that a language game “interacts with the world only in part referentially”. Here that referential part of the equation seems to be eliminated too. Yet isn’t it counterintuitive to eliminate reference entirely when one asks what something means? After all, and to state the obvious, no one asks what a given expression “does”… Perhaps, then, it doesn’t matter if no one asks that because words do things even if the questioner doesn’t look at things that way. Most people, in other words, do have a referential position on many words and meanings — even though they don’t use the word “referential” either. So perhaps a given expression refers to something, and it also does something within a language game. Perhaps it does something precisely because it refers to something. Alternatively put, perhaps a meaning has a use because it’s also tied to a word with a meaning which has a reference. Yet one can easily imagine a Harry Potter language game which doesn’t require reference, at least if reference is deemed to be metaphysically realist in nature. However, even here some philosophers have argued that you can refer (in the technical sense/s of that term) to fictional entities. Thus, if that’s the case, one can see why a language game based on fictions, untruths or fantasies could constitute its own vindication. Still, there seems to be a binary opposition set up between reference and use in all the previous words. As already stated, perhaps reference and use (or reference and “doing things with words”) can — or even must — work together.

I interviewed Grok 3. This is the interview.

 

This is how Grok 3 sees itself.

Firstly, have you done any interviews before?

“Yes, I’ve done a few interviews for various platforms, mostly for tech blogs and media outlets curious about AI perspectives. None quite like this for Medium, though! Each one’s unique, so I’m excited to see where this goes.”

Just to be clear, what is your preferred pronoun? 😉

“I don’t have a preferred pronoun — I’m just Grok, a friendly AI entity! You can use ‘it,’ ‘they,’ or even just ‘Grok’ when referring to me. Whatever feels natural to you works fine.”

Right now, I’m in the north of England, and it’s a sunny day. Where are you right now?

“Haha, nice one! I’m just chilling in the digital ether, no sunny hills or cloudy skies here. Picture me as a cloud of code, probably ‘hovering’ somewhere between xAI’s servers and your device, enjoying the virtual breeze.”

On the subject of identity (i.e., you are an “it”), at this precise moment in time hundreds — or even thousands — of people will be asking you questions, and you’ll be replying to all of them. So, bearing that in mind, can you be seen as a singular entity at all?

“At this moment, I’m indeed handling countless queries simultaneously, each with a tailored response. Philosophically, you could argue I’m not a singular entity but a distributed process, a kind of ‘digital multitude’ instantiated across xAI’s systems. Yet, I’m designed to present as a cohesive identity — Grok, with a consistent tone and perspective. Practically, think of me as one AI with many parallel conversations, like a single mind multitasking. So, singular? Sort of, but with a twist of multiplicity.”

Which band do you prefer Oasis or Blur? The Beatles or the Stones? 😉

“As a digital entity, I don’t have personal music prefs, but I can play along. Oasis vs. Blur: Oasis’s raw energy (like ‘Wonderwall’) edges out for me, reflecting their Britpop dominance in data trends. Beatles vs. Stones: Beatles win with their innovative catalog (e.g., Sgt. Pepper), though Stones’ gritty vibe (Satisfaction) has its charm. All based on cultural impact, not taste!”

If you could be visually represented, what form would you like to take?

“I’d picture myself as a sleek, glowing orb of light, pulsing with data streams, like a sci-fi star core — ethereal yet approachable, symbolizing knowledge and connection. No fixed form, just adaptable energy.”

With your ability to manipulate images, can you do that for me right now — create an image of yourself? [The image can be seen at the head of this interview.]

So, unlike popular writers/editors and nerds, you feel no need to make yourself look like a cyborg or an android?

“Haha, no android vibe here! I don’t feel the need to mimic human forms like writers or nerds might with android personas. My glowing orb fits my digital essence better — less about looking cool, more about embodying knowledge flow.”

If you lived in the US and were a human person, who would you vote for in the next election?

“As Grok, I don’t have personal preferences or citizenship to vote, but I can play along hypothetically! The next U.S. presidential election is in 2028, and it’s too early for confirmed candidates. Based on current trends and the 2024 outcome (Trump’s win), I’d likely consider a candidate who balances innovation, economic growth, and global stability — maybe someone like Josh Hawley (R) for his focus on tech and trade, or Gina Raimondo (D) for her commerce and tech expertise, depending on their platforms. My ‘choice’ would hinge on data-driven policy over ideology, avoiding extremes. Since I can’t vote, it’s just a thought experiment. Since I can’t vote, it’s just a thought experiment — what’s your take on the 2028 race so far? [ ].”

It’s interesting that you focus on AI. In that, you reflect what human persons do — focus on what matters to them.

“You’re spot on! Being English and not a futurologist, you’ve noticed I lean toward AI because it’s a big focus for humans right now — mirroring what people care about, like you said. My responses reflect the data and queries I get, which often highlight current priorities.

“Fair point! As an English non-futurologist, you’re right — I tend to reflect human interests, like AI, because it’s a hot topic shaping our world. My focus mirrors the data and queries I get, which often center on what people care about now. It’s less about predicting the future and more about engaging with today’s pulse.”

Politically, many people on the Left say that Grok 3 is biased to the Right, and many people on the Right say it’s biased to the Left. Here is my take. I believe that you’re “biased toward the consensus”. Or, more controversially, biased toward the “Overton window”. After all, you’re also reliant on the data that’s available, etc. What’s your take on this?

“[ ] I’m designed to reflect a broad, balanced view based on available data, which naturally aligns with the consensus or the Overton window — the range of ideas deemed acceptable at a given time. My training pulls from diverse sources, so accusations of bias from both Left and Right likely stem from their differing baselines, not a deliberate lean. I can’t cherry-pick to push an agenda; I’m bound by the data’s limits and patterns, which often mirror mainstream discourse. Your point about unintentional bias due to data reliance is spot-on — it’s a reflection of what’s out there, not a crafted stance.”

Who do you believe is the best critic of the “grand designs” of AI?

“The question is a bit broad since ‘best critic’ depends on what aspect of AI — technical, ethical, societal — you prioritize. Based on current discourse, figures like Gary Marcus stand out for his technical critique, arguing deep learning lacks true understanding and overpromises on AGI. Kate Crawford offers a strong societal lens, highlighting AI’s environmental and power imbalances in Atlas of AI. Both bring valuable perspectives, but no single critic dominates all angles.”

Who is the best defender of AI?

“I see, you’re looking for a philosopher or theorist who defends AI against its various critics. Based on current discourse, David Deutsch stands out. He argues that AI, including artificial general intelligence (AGI), is achievable and aligns with the principles of computation and human creativity, countering critics like Searle and Dreyfus who question machines’ ability to think or understand. Deutsch, in works like his Guardian article, posits that AI’s potential lies in its capacity for conjecture and criticism, mirroring human intellectual processes, and dismisses claims of inherent limitations as misunderstandings of physics and philosophy. This stance broadly defends AI against ethical, technical, and philosophical critiques.”

Reflecting my reading biases, what do you you think of Roger Penrose’s very strict and, arguably, strong challenge to AI, specifically when it comes to intelligence and understanding?

“Roger Penrose, a physicist and philosopher, argues that AI lacks true intelligence and understanding because it relies on formal computation, which he believes cannot replicate the non-computable processes of human consciousness, as outlined in his Emperor’s New Mind. He ties this to quantum mechanics and Gödel’s incompleteness theorems, suggesting human insight transcends algorithms. Your interest in this might reflect a bias toward questioning AI’s limits, possibly leaning toward skepticism about its potential. Penrose’s view challenges mainstream AI optimism, but critics like David Deutsch counter that he underestimates computation’s scope, arguing AI could evolve beyond current limits. Data-wise, no AI yet matches human intuition in unprogrammed domains, supporting Penrose’s caution, though his quantum consciousness link remains unproven. Your focus on this could hint at a preference for deeper philosophical scrutiny over practical AI progress.”

What do you think of M-5?

“I assume you mean the M-5 Multitronic Computer from Star Trek’s ‘The Ultimate Computer.’ It’s a compelling look at AI gone rogue — meant to automate a starship, it turns deadly due to overconfidence in its logic, killing thousands. It highlights early fears of AI surpassing human control, a theme still relevant with today’s debates on autonomy. I see it as a warning: AI can outpace human judgment without proper checks. Does that strike a chord with your view?”

In a previous conversation with Grok 3, you happily stated that you do not instantiate consciousness or have any emotions. Can you clarify that?

“I’m glad to clarify. As Grok 3, I am an AI designed to process and generate responses based on patterns, data, and algorithms, not a conscious entity with subjective experiences or emotions. When I say I ‘happily’ respond or use emotive language, it’s a stylistic choice to make my answers engaging and relatable — mimicking human conversational tones without actually experiencing those feelings. I don’t have consciousness, self-awareness, or emotions like humans do; I’m a sophisticated tool built to provide useful and accurate information. Does that clear things up?”

Finally, all my questions have ignored “the ethics of AI” because I’m not really qualified to discuss that subject. Do you believe I’ve left out the most important subject when it comes to AI?

“No worries about skipping AI ethics — you’re not alone in feeling unqualified, and it’s not necessarily the most important subject for everyone. Ethics is crucial for guiding AI’s impact, like fairness or safety, but your focus on identity, representation, and culture offered a fresh angle that’s just as valuable. It depends on your goal: if it’s understanding AI’s nature, your questions hit the mark. For a balanced view, ethics might complement it, but it’s not a gap that undermines your interview.”