So thanks Luciana for inviting me. So first thing I should say this talk is it's not the has a different title and the topic is different from the one that was initially advertised. It's not going to focus so much on the work of Robert Brandon as I hope to but more on the on on the work of Wilfred Sellers, on whom I'm writing a book. So, but nevertheless, I hope that, you know, some of the things I say or that I'll be discussing will be relevant to the central topics of the workshop. Although, I mean, I should say from the outset,
I'm not, I'm by no means, you know, an expert on computationalism and robotics or artificial intelligence. and I think Giuseppe Longo and Ioana Sait both know about those two things far more than I do. So this is more like a kind of a warm-up for their interventions, and I think that some of the things I say now will probably be fleshed out in more detail by Johan Sait in her talk. Okay, so first of all, I want to begin by focusing on the two aspects of the mind. The Renata Thesis famously said that intentionality is the mark of the mental. So in order to understand the mind, we have to characterize minds as intentional systems.
and then the puzzle consists in figuring out what exactly, how to understand intentionality. Kant is credited with a fundamental distinction between sapience and sentience. And the inferentialist project, as developed first by Sellers and then by Robert Brandon, is predicated on this fundamental Kantian distinction. These are the two aspects of the mind which map on to, you know, this distinction can be cashed out in lots of different ways. It's been currently cashed out by people like doulas such as David Chalmers in terms of distinction between the phenomenal or experiential aspects of mind, which would be sentience, and the psychological or functional aspects of mind,
which would be sapience. So sapience is reading or isciocination, whereas sentience is awareness or consciousness. and here and then there's two different kinds of intentionality okay because the intentionality of conceptual thoughts um the old hotel called the openness of thought is fundamentally different from the intentionality in sensation and awareness in other words to think about something isn't the same as to be aware of something so then we have so then the problem is articulating these three trials, sapience, sentience, and intentionality. And the question is, is the intentionality of mind rooted in sapience or in sentient? Which of these aspects will be more
fundamental? Now, phenomenology roots intentionality in sentience and argues that sapience is conditioned by sentience. In other words, consciousness, conscious experience, is what allows us to have thoughts about things. It falls that if sentience is primary, the condition for creating artificial minds would be the creation of artificial life. You'd have to create consciousness, sensory consciousness in the full-blooded sense, in order to be able to generate the kind of complex, you know, discursive, conceptual intentionality which characterizes human beings. But if sentience comes first, artificiality comes with a contradiction in turn,
as I said. Sentience cannot be algorithmically decomposed, according to, hence, the hostility amongst lots of phenomenologists or people inspired by the phenomenological tradition to the computational paradigm in philosophy of mind. And this, I guess, is exemplified by imactivism, the radical externalism, a radiante-representationalist account of the mind. So the contrary position, the position being articulated, I think developed in a way by sellers is the position which roots intentionality in sapien and argues that sentience is conditioned by sapien.
In other words, even our sensory experience, our sensations of things are conditioned by our conceptual capacities, which are characterized in terms of rule-governed discursive practice. So this is the basic tenet of inferentialism, that conceptual rationality consists of rule-governed transitions between assertions, and that there is such that even our perceptual transactions with the world are ruled government or interventually articulated. Now, what is a rule? A rule is not simply a codified regularity or a dispositional mechanism.
It's a normative injunction of the form do A and C, do A and C asterate and C asterate, et cetera. So rules are normative functions, not callable functions. And this is a crucial difference in understanding normative functionalism from classical causal functionalism. The ability to follow a rule or to do something because of a rule, cannot simply be understood as the actualization of a disposition. Deviation from a rule is error, whereas causal deviation is malfunction. Reasoners err, but nature doesn't make mistakes.
The non-actualized disposition in a physical system can always be explained in terms of the non-actuality of the relevance or the appropriate causal mechanisms that would have set it, no, actualized it. Now causal function is straightforwardly mechanizable, so therefore it's computationally tractable, and that was always why, you know, function is the kind of the Orsonauts kind of ideology for, I guess, classical computationalism. It's computationally tractable, although its adequacy is a paradigm for understanding minds have been much contested, principally by waves of anti-representational ideologues.
Another way of saying this is that causal functioning is predictable. Where prediction fails, causal dysfunction is explicable by invoking the intervention of some other causes. In other words, if some system fails to do what you expected it to do, you can straightforwardly explain you will look for some intervening cause that prevented it from doing this thing. So then the question is, is normative function computationally intractable? Is it inherently unpredictable? It's tempting to construe normative functioning's irreducibility to causal functioning as evidence of what Kant called spontaneity. The spontaneity of the understanding, which is, of course, another way of talking about freedom.
But if we want to avoid a metaphysical reification of freedom, some sort of supernatural force, we have to find a way of integrating normative functioning within causal functioning. It's not simply enough to say that normative functioning is irreducible to caudal function. Only this way may we avoid lapsing into a dualism of rules and causes or of the normative and the physical. I think it's precisely my interest in Selzer's work is motivated by the conviction that he understood the risks of generating this dualism, which is just a kind of, it's not the classical Cartesian dualism,
it's not been a classic mind-body dualism or mental-physical dualism, but insofar as a dualism is, as Brandon says, an unarticulated distinction, unless one can explain how rules interact with causes or how rules are embedded in causes, one will inevitably relapse into some kind of dualism. So whether and how normative functioning may be integrated into causal functioning depends on how we understand rule forward. Now, this is where the difficulties begin. Is it possible to follow a rule without being conscious of the rule, without understanding or intending the rule?
Either this consciousness of, if we have to be conscious of a rule, or if we have to be thinking of a rule in order to be able to follow it, then we need a rule to be able to follow the rule. Because the consciousness of the rule will be determined by another rule, which itself presupposes consciousness of it, and we're all on an infinite regress. the alternative is to insist that consciousness of the rule consists in some sort of immediate grasp of awareness of the rule as a rule but then this comes at the cost of rooting intentionalities in sentience after all because there's a kind of immediate conceptual there's a grasp of something
as something in central comprehension which is not itself articulated by a more fundamental by rule which is not itself rule government so the challenge for the influentialist is to give an account of rule following which does not presuppose some form of original awareness or consciousness of rules and And to do so, we must distinguish between understanding and awareness. Someone can understand how to follow a rule without being aware of what the rule is. This allows us to account for the difference between someone doing something in conformity with a rule, without grasping of that rule, and someone obeying a rule that they have properly grasped.
And this is really the distinction between doing something, you know, Dennett, I think, explains this very well. There's a distinction between doing something for a reason, which most animals do. Evolution generates motivations for animal behavior, sometimes extremely sophisticated and complicated motivations, reasons for doing something. But those animals are completely oblivious, need not be aware, need not know anything about these reasons, these motivating factors. So this distinction then between awareness and understanding helps us avoid the circularity
whereby consciousness thought was explained by grasping a rule. While grasping a rule is explained in terms of consciousness thought. At the same time, it allows us to explain what it is to grasp a rule without resorting to any unexplicated notion of awareness. Okay. So, now, Sellers tries to kind of, we'll try to get to grips with this topic in his discussion of language. Right. And the first thing to understand is that a game is simply, you know, a game is at once a set of rules
and a complex pattern of behavior incarnating those rules. Individual moves in the game contribute to the realization of this complex pattern. The structure of the pattern is determined by the ways in which the rules are realized through the succession of moves that constitutes this particular game. Each move in the game realises parts of the pattern according to a rule. But while the pattern is a physical phenomenon, it's tempting to maintain that the rules through which the pattern is realised are not. This is a temptation that must be resisted if we want to avoid this dualism of rules and causes.
So the task is to de-raify rule. Temptation to raify them can assume a platonic or an Aristotelian form. The platonic variant consists in raifying the rules as ideal structures subsisting independently of the pathogens that incarnate them. The Aristotelian variant consists in raifying them as ideal functions shaping inert physical substrates. to de-raify rules is to see how normative function is embedded in causal function and how logical powers are realized by causal powers. And this last specification is crucial. And I think this is really, or this I think is what I find particularly fascinating
or compelling about Sellers' approach. He insists that logical power, which is to say inferential valence, is embedded in causal efficacy. In other words, we have to understand the relationship between inference and causation. this is expressed you know this is this conviction is expressed in sellers's famous claim that espousals of principle which is to say rules are reflected in uniformities of performance
okay but what's what has to be emphasized here is that it's the espousals that are reflected in behavior, not the principles themselves. A principle is espoused through the attitude we adopt towards it. And this attitude cannot be that of intending the rule since the rule is supposed to be the source of intention. Thus, rule-governed activity cannot be separated from conditional behavior. So, now, okay, here's a quote, first quote from Sarah's about the notion of the relationship between rule-governed activity and pattern-governed
behavior. This is kind of, I apologize for the length of the court, but it's actually really, I think, quite helpful. So this is from Some Reflections on Language Games from 1954. To learn pattern-governed behavior is to become conditioned to arrange perceptible elements into patterns and to form these in turn into more complex patterns and sequences of patterns. Presumably such learning is capable of explanation stimulus-response reinforcement terms. The organism coming to respond to patterns as calls through being, among other things, awarded when it completes gappy instances of these patterns.
Pattern-governed behavior of the kind we should call linguistic involves positions and moons of the sort that would be specified by formation and transformation rules in its metagame, if it were rule-beying behavior. The key thing is we're going to try to understand this distinction between pattern-governed and rule-obeying behavior. And this highlighted the crucial claim. Thus, learning to infer, i.e. reasoning, learning to reason, learning acquiring conceptual competence, where this is purely a pattern-governed phenomenon, would be a matter of learning to respond to a pattern of one kind by forming another pattern related to it
in one of the characteristic ways specified at the level of the rule of being used of language by a transformation rule. That is, a formally stated rule of inference. Okay, so what I'm now going to do is try to give a cape, well, the absurdly kind of condensed summary of what I take to be the crucial claim in some reflections on language games, which is the interpretation of game and metagame, which is also the interpretation of pattern-governed rule-being behavior. Now, it serves me to the distinction between,
it says there's something interesting. So you can define that game. A game is defined by its rules. But the rules themselves are not part of the game. You have to know the rule or obey the rules when you're playing the game. But you don't state or declare the rules when you're playing the game. So in other words, there's a distinction of levels. And he likens this, Seiler's models, or, well, Seiler would be kind of models of the distinction of analogic language and metamanguage, even though it's not a linguistic distinction. just as in an object language an object language consists of first order you know discourse about
things, about the world a meta language is the kind of the stratum which contains the rules governing the proper use or the proper functioning of sentences in the object language so So, Sellers' paradigm is the chess. He really uses a chess analogy, very popular in the 1950s, most of the one philosophers. So here you've got these two levels. Okay, the lower level is the game level. Think about that as analogous for the object language level. And the upper level is the metagame. Every game includes its own metagame.
The metagame is a level which the rules governing that the rules that define the game find expression um so um what sellers will do is say but this transition um between levels is what happens when you um learn what you actually learn to obey a rule okay so the lower level simply did you know him you have let's say I couldn't find any symbols corresponding to chess pieces though just assume that these things correspond to a bishop and
that chain okay so a bishop shape piece of wood nets to a king shaped piece of Now, it's important that you can't use any expression such as bishop or king is in a way is going to be a metagame expression because you need to know the rules to understand what it is for something to be a bishop or a king in this game. Seller's distinguishes between perception, reasoning, and action. So he says that the initial, just as in everyday discourse, we perceive a state of affairs.
We perceive a water bottle sitting on a table. And then this perception may, we see that the bottle of water is on the table, and then we say the bottle of water is on the table. This is a language entry transition. This is when you enter from a non-linguistic state into a linguistic state. There's a similar transition from the game to the metagame. There's a language entry transition from perceiving a configuration of bishop and king-shaped pieces of wood on a board to thinking that my king is chained by his bishop.
So there's a transition from the game into the metagame. There's an ascension into a higher level where you think you have a conceptualized thought, which is about these pieces of wood that you are currently perceiving. This is the first step, the moving from the game to a metagame. The next step is infroom, moving within the metagame. This is an interlanguage game. Any instance of reasoning, any claim such as red is a color, crimson is a shade of red, these are interlanguage moves because you don't need any information about the world.
if you don't rely on perceptual information, to be able to execute these inferences. So from the basic, the rule of the game, if a bishop checks my king interposes a pawn, I make the move to then I must interpose a pawn. Nothing has yet happened. I haven't moved any pieces of wood yet. The final stage, the third and final stage, is the language exit transition, which of course wants the action, where you have your multilateses by the injunction, interpose a pawn, okay, which makes you do something. You move a pawn-shaped piece between a bishop-shaped piece and a king-shaped piece, okay.
Now these, okay, this is a very pretty cartoon version of King Sellers's argument, okay, but But the claim is that, the key claim is that all that is required in order to carry out, in order to reason, in order to do something for a reason, or to do something because of a rule, as opposed to merely in conformity with a rule, is a kind of practical know-how, and not theoretical knowledge, not knowing that. So I think this took me a long time to understand this. But the way in which you avoid the infinite regress of whereby, you know, if you need a concept for rule,
how can you understand what a concept is without having a rule to follow the rule? If you need it, either you have either your ability to recognize the concept presupposes that you already have another concept okay and then you have this infinite regress you need to have a concept of a concept of a concept or a rule for a rule for a rule or if you terminate the regress by saying no you just have this kind of intuitive you know pre-conceptual awareness which allows you to simply to understand something as something but then we're back into what selich calls the myth of the given the idea that things you could have direct preconceptual acquaintance with something as something
that's one version of the myth of the game and you sell us turn things sold a problem by saying the transition from language to mental language or from game to mitigate is itself a kind of know-how and not knowing that you need to know how to ascend from one level to another. And even at the meta level, at the metagame level, all that is ensuring or rather kind of precipitating the transition is practical no-vow. So that reasoning, inference, is just a kind of practical competence
and not necessarily a kind of theoretical reflection or deliberation. So perceiving a specific configuration of bishop and king-shaped pieces of wood as a bishop, so perceiving, sorry, a bishop and king-shaped pieces of wood as a bishop checking a king, inferring that if one king is threatened by a bishop, inter-bordafone, and interfosing one's pawn are all rule-governed practical competences. So reasoning is essentially a kind of practice
which supervenes on certain habituation or some kind of basic behavioral condition. You have to be trained to reason, but that training does you with the capacity to effectuate what seem to be spontaneous transitions as opposed to kind of mechanically ensure causal transition. Here's another quote. This, I think, will be the last one. From the same paper. Words which mention the positions of a game, which is to say position words, are, we might say, the observation words of a rule language.
And in addition to their syntactical role in the rule language, they occur in sentences which come to be occupied as a result of a language entry transition into the rule language, in which transition the stimulus is a situation of the kind meant by the position words. Action in joining contexts, on the other hand, are the motivating expressions of the rule language. And in addition to their syntactical role in this rule language, they occur in sentences, the occupying of which is the stimulus for a language departure transition out of the rule language for response, which is an action of the kind mentioned in the motivating context. This is basically Sellers' gloss on the schema I just tried to show.
So the metagame then states the rules governing the game. Because competence in the game requires competence in the metagame. The ability to see bishop, horn, and king-shaped objects as pieces in the game is a rule presupposing familiarity with the other rules governing relations between pieces. so in this sense i think it's worth it to say that the meta game states explicitly the rules that are implicit in every movement and position of pieces in the game you know hence this is why when you're playing a game you don't need to state or to conceptually articulate
the rules in accordance with which you are playing the game. Moves made because of the rules, and this simply means to make a move because of a rule, which is to say that's the only sense in which you can be sensibly playing the game, or acting for a reason, are unpredictable because they cannot be tracked by facts at the level of the object game, or at least this is a conjecture. One way of understanding the unpredictability of sapient beings, of reasoners to themselves, is because there are two, there's a stratification of levels, and information about the object level will not suffice to help you understand what is going on at the meta game level.
Now, this is not to say, or at least it would be very hasty to conclude from this, therefore, that reading, or what we call the spontaneity of the understanding, is somehow radically indeterminable or uncomputable. So I think we'd have to be hesitant about simply saying that this is a way of understanding freedom without any kind of metaphysical libertarianism, without postulating some kind of numeral will. Okay. Now, but language must enable language users to find their way around in the world and satisfy their needs.
So if inferential moves make a difference in the world, like reasoning makes a difference in the world, language must be articulated with the world otherwise than by representing it. Because we can, you know, this inferentialism, you know, sellers and random insist that language does not represent, or meaning rather, meaning doesn't stand for anything in the world. It doesn't designate features or properties of the world. So then the question is, how do the rules governing perception, reasoning, and action gain traction on the world? Okay, and here now I'm going to have a quick discussion of another paper by Sellers called
Knowing and Being Known, in which there is a fascinating discussion of whether or not the reasoning, rationality as we characterize it, could be mechanically instantiated. But what I want to do, the point about this is to understand the connection between, on the one hand, the discussion of the relationship between games and metagames, which is the relationship between pattern-governed and rule-obeying behavior, and the, let's say, the kind of the modeling, the kind of modeling of the worlds,
that's organisms and other kind of, you know, let's say kind of sentient systems need to be able to engage in in order to be able to kind of find their way around and do what it is they want to do. So sellers, this is, you know, paper written in the early 60s, so it probably, its frameworks may seem anachronistic, but I think the basic point is still very powerful. The robots has a wiring diagram which determines transformations from senses to other senses in accordance with mathematical and logical principles. So these would simply be kind of logical inferences
of transformation principles. If it's trying to find its way about in the world, in our world, we assume that it would have to have something analogous to a principle of induction. So it would contain the equivalent of inductive generalizations such that if its tape contains sense fears such as lightning at ping, lightning at place, place one, lightning at time, chi one, thunder at place plus delta P T plus delta T. And no sentence fares like lightning at P T, quiet at P plus delta P T plus delta T. Then it would print sentences such as
whenever lightning at P T, thunder at P plus delta P T plus delta T. so this is through this series of inscriptions okay so what is unfolding through the wiring diagram and the instructions that generate these in so this series of inscriptions in the robot for wiring diagram is a representation of a representation or a model if you will of the robots world its environment. So, in a way, all organisms need some kind of representational system in order to be able to map the world around them, and this is a function of representation.
So then the crucial decision then becomes between signification, which is semantic and non-govern and properly conceptual and representation representation or mapping so there were just one thing to know things about the world and to be able to think about the world and it's another thing to be able to find your way about the world on the basis of what you think but you need two different accounts to articulate these levels in the order of significant in the in the conceptual dimension, or the order of signification, let's say the inscription, this inscription on the tape out of the four dogs
can be said to signify lightning. And the pattern, you know, the pattern we see in square belts 915 can be said to signify lightning at place 9 at time 15. What is it to say that something signifies? Signification here just means meaning. If you say that this linguistic expression means this, according to Sellers, what you're doing is you're establishing a functional equivalence between a quoted expression in a foreign language or some kind of unfamiliar linguistic expression
and a familiar expression of your own language. In other words, you're pointing to this sign or inscription in this unfamiliar, you know, tongue and saying it plays the same role in this symbolic system as this expression here with which we assume our interlocutor is already familiar. So in order to say what something signifies, you have to show what an expression in your home language does. So, for instance, if you say that, you know, if you say that éclair in French means like me in French,
someone you can, you know, the point of that, that's an informative claim insofar as someone who doesn't understand French you're taking this sound or this inscription in French, which your auditor doesn't understand, and you're saying it plays exactly the same role in French as lightning does when I say it and mean it in English. So this is what it means to establish a functional equivalence between the robot-y sign design and the English sign design, likely, as well as one between this other robot, he's signed design, and the English expression of lightning at place 9 in time 15. So this maps, in a way, picture,
a map is an orientation system which allows a system to orient itself in the world by picturing, picturing certain features or aspects of the world. Okay, again, this is one concept I think which has been much misunderstood and I think unfairly kind of dismissed. But I think it's particularly helpful if one wants to kind of show, In fact, it's indispensable if one wants to understand how rules are embedded in the causal order. Picturing does not consist in a relation of resemblance between representation and represented.
It consists in the structural equivalence between properties of relations among representations considered as natural objects and properties among represented objects. So, there's a systematic correlation between tokenings of the robotis, lightning, éclair and lightning, in robotis, French and English, and instances of lightning in the world. But the points, again, this is, I think, a crucial nuance. The point is that Sellers describes picturing as a second-order isomorphism. It's not an isomorphism between two objects.
It's an isomorphism between two systems of objects, between relations, as a system of objects consists of a system of relation. To call picturing a second-order isomorphism is to say that it's a relation of relation. So there are matter-of-factual, and these relations can be understood as matter-of-factual properties relating particular occurrences of these inscriptions and vocalizations to particular occurrences of lightning. This system of relations constitutes pattern in the causal order, and it is this pattern which incarnates the rule. Okay, and now I think, and actually what I want to kind of conclude with is, I think, a possible contrast between patterns and processes.
I think Johanna will have much more to say about this this afternoon. But I think I'm increasingly convinced that there is a really significant distinction between patterns and processes in the seller of his work. A pattern is a complex mosaic of matter-of-factual relations. but if Sellers insists the world consists of things and not fact then patterns are not processes at least insofar as Sellers seems to be committed to a process ontology what he calls absolute processes
I'm assuming here that absolute processes don't have propositional form Maybe that's wrong. I don't know. I might be wrong about this. But patterns, insofar as they're composed of object-bound relations, seem to have, you know, they're statable. You can describe a pattern in terms of a series of facts, you know, the facts which instantiate this pattern. And these facts are object-bound in that they consist of relations amongst objects, amongst well-individuated objects in space and time.
And it seems that, or at least, you know, it's not obvious that processes can be, what Salazar just says, that processes are not object-bound. absolute processes cannot be attached or they're limited in terms of objects or relations of objects. Okay. Well, that's... I think I'll stop. I'll just stop here. Thank you. Okay, so first, thank you to all the speakers for accepting to come and speak today,
and thank you, Ray, for this excellent talk. I wanted to pick up on this final kind of point that you made, actually, because it relates to something I've been thinking about for a while and my research, which is based on noise. And so the distinction you made there just now between the pattern and process is something I've been thinking about. Because if we think about, well, a pattern can be thought, can be understood as a regularity. And even if it's a very complex one, like you're saying here in this kind of mosaic of of matter of factual relations um uh whereas
we could think of uh uh you know non-periodicities in in uh in um you know material relations between things which would be kind of uh uh perhaps not so uh not statable as facts in the same in the same way um so uh i you know i wonder whether kind of you know we could say at the at the limit you know uh a of uh of of the this of calling something a pattern there must be something which is not pattern which would therefore just be processed uh and uh you know which couldn't be decomposed into a set of regularities. So then I thought, I mean, one of the things I would like to kind of just to situate where
our speakers here today in some way, if I may, which is, I think there's a kind of a broad agreement amongst the speakers here with that for example there's uh uh uh they all they all reject a kind of simple computationalist picture of the of cognition uh and uh and you know or and move beyond its description in terms of mechanistic causality um so but there's a there's I think a range of ways that that has taken by the different speakers here today. And I think so Giuseppe Longo has written very strongly against the computationalist paradigm
and has drawn on the fact of multiscale randomness at many layers in order to do so. And so he's arguing that there is, for example, in processes like biological evolution or economic systems, the processes are so complex that they cannot be given entailing laws. So they can't be modelled as an NK landscape, as an NP-complete problem, for example. So they're not algorithmically searchable or definable.
Now, on the other side, I think Johanna Seibt and Ray, to some degree, are arguing that there is not a necessary fundamental limits to the computational possibilities of decomposing those functions. functions so that there so that it is possible to to uh to to decompose those inferential strategies uh and to to automate them in some senses uh now um in in some way this uh this you know well let's take uh let's take um uh an example of uh you know of course uh kind of
biological biological um systems such as a fish its movement is unpredictable uh but we we can we can still fish catch fish right because they're they've got a regularity to their movement right they've got pattern governed behavior which we can regularly rely on you know with nets and and and and so on but uh but when it comes to uh complex systems which involving higher high cognitive animals then then it's a much more different different problem uh and involves um you know for example um the performativity of the model the counter counterperformative counterperformativities and a complex kind of uh array of these effects of
of the of the inferential practices involved in in in that in that process so then uh i think you know when it when it comes to uh to kind of the the uh the problems of prediction today and the way that they are they're being unfolded in in say uh uh aIs on on the on the internet um there's uh there's a um there's a real uh it's really well as as you said palongar argues this is a frame problem rather than an NP-complete problem.
So it is a matter of defining the context within which a probability distribution unfolds rather than searching an already configured space. So the space, because the space of possibilities are constantly changing and constantly in flux, the parameters are constantly changing, then we can't say, we can't give a full list of all the possibilities of the states. So what we need is to define the context for the operation of that. And I think that's exactly what's happening in rule-governs behavior rather than pattern,
sorry, rule-obeying behavior rather than pattern-governed behavior. So it's the definition of a context for action, which would be effectively a solution to the frame problem. So, yeah, in conclusion, I'd just like to kind of bring up the questions of randomness, contingency and noise with regards to the main items of debate today, which are prediction, process and reason. And I hope that we can articulate some kind of relations between them as we go throughout the day. Thank you. Thank you.
Do you not respond? Do we have maybe 10 minutes for a question? You would like to go for a question. Okay. Do you mind? Okay. One there, so... What's your name? David. And then... Oh, God. There you go. Sorry for recording, Bram, thanks. I was interested in Nico's comment about computational tractability, as well as we're kind of rewiring sellers via Brownton. I feel one of the things that Bray took from Davidson is the point that no state is just going to have an inferential role.
is still in this inferential world going to depend on all sorts of auxiliary commitments. You know, what we believe about the world will determine what we can inferred for other people. So you need some kind of update algorithm, if you like, whenever you're dealing with other speakers. And of course, that could be really complicated because how do you figure out what kind of new synodary between other people, even what you've already attributed to. And that could actually be a really complex process. And it's not clear that that's a track. Obviously, an algorithm could be computable but not track. You know, arranging all the permutations of further things
is perfect. You can state the algorithm. You know, there are just too many possibilities to actually... I guess there is a typical computational tract that would have given that closed migrating a kind of function synaptics. Okay, so I mean, when you say it's not algorithmically tractable, so it wouldn't be possible then to articulate the inference and something like a kind of, Yes, a kind of reinforced, something that is kind of simply algorithmically, yes. And I think there's an issue there simply because of the complexity of internet putting up.
Yes, sorry. Yeah, that's all. I don't know about it. I know it can be. in Brandon somewhere goes through a sort of describes a procedure that the speaker has to adopt in imputing commitments to other speakers and even certain inferences on the faces of other beliefs they have it's pretty corporately and it's not clear whether we could write a program to do that so I guess that security is under the problem yes I mean I think look I mean I think what is
powerful I think or still kind of powerful about this I found is that or at least you know the Szilagian version that I find quite compelling is that the you know the pattern-governed disposition that, you know, speakers need to acquire in order to be able to think, you know, to speak and think, are not, you know, they are just kind of, they are a set of cultural contingencies, okay? They're going to be contingent because of, you know, biological, historical, and cultural circumstances.
So there's no claim that these... In other words, these are only kind of, I guess, and you can correct me about this if I'm wrong, but I guess these will only be algorithmically decomposable in some kind of trivial sense in which anything is. So they're not kind of... because of the primacy of material inference in this account, what generates a material inference is simply a habit, a kind of acquiring a habit. But if we're going to make sense of our speakers,
that's going to have to be acute upon sex in Sicilian. He goes, yes, I just think Mrs. Maladroal, so this is a nice arrangement of epithets, you've got a kind of interpretive meaning in the light of the situation you're in, rather than in terms of some standard kind of rule. Ah, but I think, okay, this is an important, I think once you understand rules, I mean, rules are context. This is why, in a way, the ability to understand a rule is not necessarily itself conceptually tractable. You can't break it down into a series of instructions. The whole point about normative functionalism is that rules can't be decomposed into a set of propositional statements because it's know-how.
It's a kind of know-how. So it's highly context-sensitive and context-dependent. This is why it's like an old riding a bike example. So knowing how to reason, knowing what to say is like riding a bike in that it's an incredibly... It's a kind of accumulation of an incredibly complicated set of kind of if, you know, circumstances. But although when, you know, these philosophers give example, they always say as if, you know, you kind of exemplify rule propositionally by saying in circumstance, when in circumstance, C do A. but it's not clear in fact whether you know rules are um can be propositionally encapsulated and
the whole point is that you don't need to rehearse a series of propositions to follow a rule that's the basic claim and so in that sense i think it's it's not so kind of you know it's kind of i think it kind of recognizes the same difficulty with classic kind of, you know, rule-based accounts. Well, thank you, Corey. I think you, we have been presenting the decision that herein, the theory, language, language in a very effective way. And this is a very critical distinction in mathematics and mathematics. It clearly helped work, but in no way is intrinsic. it is objective, it is strongly contextual, and it, both very often in the Khrushchev cases,
totally disappears. And in Herbert first, in 904, Tolti had proved the consistency of arithmetic with May conjecture in 1900 by a group by induction. Therefore, in Karif Tolti, well, you know, Tolti are proving the consistency of arithmetic by induction, which is the fundamental axiom of arithmetic. So, super right. Then, Herbert Tolti, I do that. In the meta theory, the test learned the notion was invented. He said, I'm going to prove by meta-induction the consistency of arithmetic which contains induction. Then Herman Baal told me, what do you know, Professor Sir Pilbert, who was a student at the science, was very specific. He, Mecca-theoretic induction is actually theomatic induction, because in the same mathematical stuff.
And indeed, this is what Gödel did. What he did is that he showed that the metatheory of Arrhythmia, which contains the notion of Prouba Bede, which is a metatheorethal notion, can be fully encoded in the theory. So the metatheory by Gödel's proof is a subset of the theory. Anytime you have a formulation, anytime you have a formal theory in mathematics that includes arithmetic, this theory can encode and thus contain as a subset its timetistic metateory that's the core of the builder's book. So as soon as you have finite of descriptions and your theory metatreat in finite of descriptions and your theory is powerful enough to contain arithmetic, the distinction totally disappears.
As soon as your rules are written with the positional finetistic action, they belong to the theory by encoding. So it's a very effective difference to say, for example, I write my notes in Italian as meta-directed for Mars on my English right, first strike on B for distinction. But you must be aware that it totally disappears by any word at the formal level. And in other contexts, they may have different meanings. Okay. That's not okay. Oh, I thought it was the question. No, that was the question. Well, okay, one thing I'll say...
Put it on your mark. Two things. One is that this, you know, as I understand them, Selich is not... I mean, the whole point is that this... In a way, there's a kind of a boundary between levels seems to be porous in some way. It's not clear insofar as in a way he's saying that the ability, I mean, whether you want to call it, you know, maybe it's misleading to characterize it in terms of, like, object and, you know, object meta, you know, first order, second order. Because, in fact, I think what he's interested in is a kind of reflexivity or a kind of involution where patterns, I guess, certain patterns become, I guess, kind of embedded and nested within one another.
Now, I don't know if that's, I mean, the work that's done by the distinction is explaining how you can treat something on one stratum of language as if it was a perceptual object, as a mere kind of inscription in terms of its kind of perceptible properties. And then there's a kind of a, you know, the use mentioned distinction also does some work in this argument. Now, okay, so I mean, do you think therefore that you think that every distinction,
simply, do you think that reflexively as such is beholden to this you know, to this kind of bad kind of stratification that you think is redundant? No, on that side I think it's been technically very useful and keeps being useful all the time, but be must be aware that according to the context there is no section external role of the meta theory with respect to the theory. As soon as you are infinitely sticking with the things, and as soon as your theory is powerful enough, the meta theory may be included in, may be encoded in. As I often say, it breathes reflexivity within the theory. That's the trick of Godot.
Exactly. And then he writes a formula saying, I am not provable. He is encoded in arithmetic, and then it is not provable. but the notion provoking is metatrading, so indeed he enforces the reflexivity he brings at one level. The power of reflexivity due to the existence of pretendly the existence of two others is even stronger of them. But we handle this, believing that it is quite normal, to reflect for ourselves to access hope of sobriety. Life is all of that, and in Norvins almost everything is correlated the topo-lmose-evilene, where it's in the global level which is distinct from the local level. The local level is causally related, intricate to the global level. These sorts of flexibilities
are everywhere, but we cannot detangle them in an artificial-led way by saying, for example, in biology, it's totally extinct, the molecular-led, from the urbanism-led. No, because there is no when I correct ask me that is causing independent from the group of that in best person so we have to face this reflexive insult the time okay okay thanks a question John perhaps more also come a question and I think is this thing and so it made So, from this understanding, in the following sense, mathematical objects have identity
criteria that are very well defined, functional equivalence, so very strict equivalence of functional role. But Semer's trick of a distinction between the object language and the meta-language is of course precisely in order to point out that the inferential role of anything that is at the object hand, which is different from the influential role of the same icon words that is articulated at the mezzanichistic level. So there's simply no identity of the functionality of the two expressions. And I see Salas very much in the story that I mentioned, I don't know if they reviewed them further then. One must not forget that these texts were written at 1953.
And at that time, developmental psychology was on the rise and people discovered that it takes children some time to acquire the normative discourse. So this is a kind of philosophical theory of language. It is very close to empirical research in developmental psychology and is opposed to bring out the interest in fact that we first acquire our everyday observational language. And only later in life, we acquire normative discourse, modal discourse, mental discourse. And Seller simply wanted to make the point that you stressed. that we never get out of the pattern-downed behavior.
Practice is what it all remains, even though we gain the illusory impression that we are talking about qualia and conscious state and mental states and thoughts as something that is very different from anything that could apply to reacquire a pattern-downed behavior. So I think my question, if I should formulate this into a question, is that the notion of a meta game, that is something that you just said that in relation to the meta language. I mean, the relationship between game and messaging and all defining.
yes I it took me I mean I spent years trying to understand this distinction and I'm still you know I'm far from confident that I understand it I mean he introduces the kind of I mean he begins by saying that what seems to motivate the distinction initially is to say that there's a difference between that on the one hand, a game is defined by its rules. That's a minimal kind of... What individuates a game is the rules. And although every move playing the game
is obviously kind of determined by those rules, the rules themselves are not part of the game. The rules themselves are not part of the game in the sense in which the objects with which you play the game are. So, for instance, when you play football, you play with, like, there's grass there, there's men in shorts running around, a white ball, goalposts, but those are all the objects. There's nothing, you know, they're playing football, but the rules of football are not kind of you know part of the apparatus that is being manipulated or that is in the game um i mean so i think that that i mean as i understand it that's what
sero seems wants to be getting at um so maybe i mean i think maybe the the distinction is unhelpful because i think he ends up saying it's almost as if you know the distinction is it's not really maybe it's unhelpful maybe it's actually a kind of an unhelpful distinction because it encourages you to think that encourages one to think that there is this in this hierarchy and it's not clear that it is it's not clear and because well as you say that there is a difference between the functional rule of expressions at one level and its functional rule at another. But then maybe talking about it in terms of the kind of levels,
maybe it's context more than levels, but I don't know. I struggle to understand that distinction, and I can get a sense of what Sarah's just trying to get from it, but I think well obviously it leads to if you reify it and kind of turn it into some kind of fundamental distinction then it seems to me all sorts of difficulties would ensue Okay, thank you very much I think we have a break of 15 minutes to have a coffee and then come back for this episode, thank you very much