Welcome back to Awakening from the Meaning Crisis. This is episode 29. So last time I went through with you a series of arguments, trying to show you the centrality of the issue of Relevance Realisation. I want to review that with you and then try and begin an account of how we might come up with a naturalistic explanation of Relevance Realisation, and then build that into an overall plausibility argument about using that notion of Relevance Realisation to explain many of the features that we consider central to human spirituality, meaning making, self-transcendence, altered States of consciousness and wisdom. Before I begin that, I want to remind everybody of how much the work I'm talking about now has been done in collaboration with other people, especially the work with Tim Lillicrap and Blake Richards in 2012, the article we published in the Journal of Logic and Computation. Work with Leo Ferraro in 2013. Some current work with Leo Ferraro, Anderson Todd, Richard Woo. Current work I'm doing with Christopher Mastropietro and some past work with Zachary Irving and Leo Ferraro on the nature of intelligence.
So, we want to take a look at what we did last time. We did, very quickly to remind you, a series of arguments. A series of arguments that pointed towards how central Relevance Realisation is. We did arguments around the nature of problem solving. And you remember, we saw the idea there of the Search Space, as proposed by Newell and Simon, and we faced a couple of important issues there. We faced issues of Combinatorial Explosion and what we need is Problem Formulation or Problem Framing that allows us to avoid combinatorial explosion by zeroing in on relevant information. I also proposed to you, and I'll return to this later, that problem solving is our best way of trying to understand what we mean by intelligence - your capacity as being a General Problem Solver. Also, we know the problem of Ill-Definedness - very often a problem formulation is needed in order to determine what the relevant information is and what the relevant structure of that information is. So that, again, points us into relevance. These two together (Combinatorial Explosion & ill-definedness) also pointed towards a phenomenon we've already talked about — insight — and the fact that you often have to solve a problem by altering your problem formulation and re-determining what you consider relevant.
We then took a look at Categorisation. I'll come back to this again in another way, a little bit later in this lecture, but we took a look at how categorisation ultimately depends on judgments of similarity and can get into an equivocation there. We can equivocate between a purely logical notion of similarity in which case any two objects are indefinitely similar or dissimilar to each other. And if we mean, instead of logical similarity which would not help us to categorize, psychological similarity? Okay, then we're talking about making a comparison of two things in terms of the relevant features of comparison, the relevant aspects. So we're into relevance and we're also introducing an important idea [that] I want you to remember; this notion of an Aspect, a set of relevant features that cohere together and are relevant to us, especially in projects like categorisation. So we keep getting [Relevance].
Of course, if you remember, doing good Cog-Psi I do a convergence argument (draws Convergence diagram again) to get a trustworthy problem or construct. And then I basically do a divergence argument (draws equally balancing, diverging lines on the right) to show how it has the potential to explain many important phenomena and establish a relevant balance between them. And so that's what I'm building here (indicates the balanced construct drawn on the board). Right now we're on this side (left; converging), how all these things are converging on Relevance Realisation (writes RR in the circle in the middle of the diagram). And then, as I said, can we use this to explain many of the features that seem to be central to human spirituality, meaning making, self-transcendence, altered States of consciousness and wisdom.
We then took a look at Communication. [-] We might've done the robot first. It doesn't matter! We did communication and we saw the issue there is the fact that you have to convey more than you can say. And then that led us into the work of Grice and the series of maxims that make conversational implicature possible. And that got us into that all of the maxims collapsed to the maximum of being relevant. We then did [-] the issue of robotics, the actual interaction with the environment. Here's the idea of being an agent and we saw the robot was trying to pull the battery that's on the wagon and that wagon also has the bomb on it. And what we saw is the problem of the Proliferation of Side Effects. You can't ignore all side effects or you'll be grotesquely stupid. You can't check all side effects or you'll be grotesquely incapable, and so therefore you have to zoom in on their relevant side effects. So again and again, and again everything is centring on this (indicates the diagram of convergence/divergence construct - Synoptic Integration of Relevance Realisation).
I want you to also now remember a couple of other things from previous lectures, how we talked about the convergence argument. This is an independent convergence argument when we talked about consciousness — not the nature of consciousness, but the function of consciousness — all the convergence arguments (duplicates the convergence diagram for consciousness), that what's going on in consciousness is doing Relevance Realisation (puts RR at the centre of this convergence diagram for Consciousness). Especially in complex, ill-defined situations in which our agency is directly involved. So consciousness seems to be bound up with Relevance Realisation. And we also talked about how this overlaps with how Working Memory -- the work of Lynn Hasher -- the job of working memory is to be a relevance filter and to screen off irrelevant information and allow in to processing, deeper processing, more relevant information. And I also pointed out — I want you to see how all these connections are forming — that there's deep connections between working memory and your measures of your general intelligence - how intelligent you are. So we see that we're getting, actually, a very powerful convergence argument towards the centrality of Relevance Realisation as constitutive, as constitutive of your intelligence, your cognitive agency, as significantly contributory towards your existence as a conscious being.
And I also suggested to you last time that this notion of Relevance Realisation — and this is what we're going to develop today — may be a way of explaining that sort of fundamental aspect of meaning — the kind of meaning that was lost in the Meaning Crisis that's expressed in the three orders, in which we were pursuing coherence and significance and purpose — that sense of connectedness, connectedness. And I'm going to try to argue that as we understand what relevance is, that relevance is exactly that sense of connectedness. So there will be deep connections between meaning and relevance (from boiling down all the arguments to being about relevance). There's deep connections between relevance and agency. That's the whole point about the robot (robotics) and communicating (communication). And there's going to be deep connections, we've already seen, between meaning and agency (completes an important triangle on the board between Meaning — Relevance — Agency), that one of the whole things about agency is its relationship to the arena, the agent-arena relationship, and how that grounds, that's the meta-meaning grounding of all our other more specific meaning making projects.
So I hope I've made at least a good convergence argument for you. That many things converge upon, many things that we're interested in — many central, defining features of intelligence and agency and aspects of the functionality of our consciousness — everything is sort of converging on this Relevance Realisation. What I want to try and show you now is how you might move towards — and this has been sort of the core of my, I guess you'd call it my scientific work — how you move towards trying to offer a scientific explanation of relevance and what that would look like and the difficulties you face doing so. I also want to try and argue that there's good reason to believe that we're talking about a unified phenomenon, a unified thing here: relevance. That this isn't just a family resemblance term for a lot of disconnected things, that there's reason to believe this is a central thing.
Let's start with trying to offer theories of relevance, and there are good ones out there. There's the work of Sperber and Wilson and others, and I will refer to some of that work as we move along, but let's try and work towards it at the metal-level. What do we need for a good theory of relevance to do? What kind of mistakes do we need to avoid when we're trying to explain relevance? The main mistake that I want to point to is a mistake in which we are arguing in a circle. If you remember, this is part of what goes into things like the Homuncular Fallacy - remember when I tried to explain vision with the little man in my head having vision?! I don't want to use whatever I'm using.../ Let's put it this way: whatever process or entity that I'm trying to use to explain relevance should not itself require relevance. What do I mean by that? If I have something "X" and I'm using ["X"] to explain relevance, [then] [X] cannot itself presuppose relevance for its function, because if it does that, I'm ultimately arguing in a circle. I have to find processes that are themselves, not processes that realize relevance if I'm going to explain in terms of those processes, Relevance Realisation itself.
Another way of putting this is [that] I ultimately want to explain intelligence in terms of processes that are not themselves intelligent. Because if I don't, if I'm always explaining intelligence in terms of processes that are themselves intelligent, that is no different than homuncular fallacy of explaining vision in terms of internal processes that are themselves visual processes. So that's going to be a guiding methodological principle. Now that turns out to be very powerful and as many people have pointed out — Fodor famously has pointed out in repeated places — it's actually very difficult to explain relevance without presupposing relevance in the machinery that you're using to explain it.
Let's take a look at some candidates. We might think that we could explain relevance in terms of how we use Representations. This is a very powerful way we think about the mind, that there are things in the mind, ideas, pictures, that stand for, represent the world in some way. We might think that perhaps it's much more that relevance is a function of Computation; computational processes. Or we might think that we explain relevance in terms of what's called Modularity, that there's a specific area of the brain dedicated to processing relevance. (writes Representation, Computation and Modularity on the board.) I want to take a look at each one of those, and I want to try and argue as to why I think they're inadequate and what that helps us to see. And what I want you to see is, and I'll try to show this along the way, that [-] if, and I'm trying to make it more than an if, but if Relevance Realisation is so central to our meaning making, our cognition and our consciousness and our self-transcendence et cetera, as we learn about how we have to best try to explain or understand it we should garner lessons about how to best think about and reflect upon human spirituality. At least in the terms that I have defined it for us.
So, Representation... Now this is just a terrifically hot issue both in terms of interest and controversy within cognitive science in general and I'm not going to try and completely decide this issue right now although I think I'll say things that are pertinent to that debate, but let's take it that what we mean by a representation is something, as I said, some mental entity that stands for, refers, directs us towards an object in the world. That's all I need! Whatever else representations are in all that controversy, that’s all I need for the point I want to make! Because I want to show you something very important about a representation and I mentioned it a few minutes ago and this is a point that John Searle has famously made. Representations are aspectual. OK so I hold this thing up (pen) and you form a representation of it. Remember all the things we talked about when we talked about categorisation, we talked about similarity etc. So when you form a representation, you do not grasp all of the true properties of this object because all of the true properties, the number, is combinatorially explosive. We’ve already seen that. So out of all the properties (draws a circle) you just select some subset (draws a small wedge of the circle), and what subsets do you pick? Well, you pick a subset that is, here it comes, relevant to you!
Are they just a feature list? No, we’ve already seen that along time ago; they have a structural functional organization, they are made relevant to each other.
So here’s what we’ve got: a set of features that are relevant to each other and then a set of features that have been structurally functionally organised so that they have co-relevance, is then relevant to me. That’s what an aspect is (underlines aspect in aspectual). So whenever I’m representing anything, this is a marker (holds up marker), however I could change it’s aspectuality (changes his grip on the marker): it’s now a weapon! And we do that all the time! In fact one of the ways we check peoples creativity is to do exactly that; we will give some object and say how many different ways can you use it? How many different ways can you categorize it? Namely, how many different ways, how flexible are you at getting different aspects from the same object? So representations are inherently aspectual, but notice the language I’m using: You’re zeroing in on relevant properties out of all the possible properties, you’re structuring them as how so-relevant to each other and then how that structural functional organization is relevant to you. Aspectuality deeply presupposes your ability to zero in on relevance, to do Relevance Realisation. That means that representations can’t ultimately be the generators, creators of relevance, they can’t be the causal origin of relevance. Now, can representations feedback and alter what we find relevant? Of course, nobody’s denying that. That’s of course why we use representations! But [what we can’t serve], they can’t serve as the ontological basis the stuff in reality that we’re trying to use to generate a noncircular account of Relevance Realisation.
Now that’s going to tell us something really interesting. It’s going to tell us that if this meaning and this spirituality is bound to Relevance Realisation, that the place to look for it is not going to be found at the level of our representational cognition, the level of our cognition that is using ideas, propositions, pictures, etc. Once again I am not saying that those things do not contribute or affect what we consider relevant. What I am saying is that they are not the source, the locus of how we do Relevance Realisation. I want to show you have this cashes out even in an empirical manner. This goes to some really interesting work done by Zenon Pylyshyn on what is called Multiple Object Tracking. Multiple Object Tracking is really interesting. So basically what you do is give people a bunch of objects on a computer screen, let’s say I have x’s and o’s (draws several on the board) and they’ll be different colours and different shapes all kinds of different things like this, and what I do is I have the objects move around and let’s say this was a red X and then after it moves around I ask you “where is the red X?” and you have to points at it. I may ask you “where is the green circle?”, “where is the blue square?”, you get the task… Now what’s interesting is how much you can do this! You can track about eight, that’s on average, objects reliably. What’s really interesting about them is the more objects you track the less and less features you can attribute to each object. What do I mean by that? Suppose I’m tracking — well that’s six shapes (on the board) — suppose I was tracking the red X and I have to keep it… I can, after lots of movement, say “oh, it’s there now. It started there, and it’s there now!”. What I won’t notice during that is that the red X has become, for example, a blue square! So all of its content properties get lost! All I’m tracking — and I need you to remember this — is what you might call the hereness, where is it, and the nowness. Where is it?
It’s here now, it’s here now, it’s here now, it’s here now, it’s here now (pointing in various different directions in the room). Everything else, it’s shape, it’s colour, it’s categorical identity, all get lost! So he calls this FINSTING. This stands for Fingers of Instantiation. it’s basic idea is like this: your mind has something equivalent to putting your finger on something - I don’t know what this is (water bottle), suppose I didn’t know what it was, I put my finger on it. I don’t know what it is, I just know it’s here, nowness! And it’s here now, it’s here now… (moves bottle around, with finger attached!). Here and now are indexicals: these are just terms that refer to the context of the speaker, so here now (lifts the bottle and moved it around with his finger again) so it’s here now and it moves around and my mind can keep in touch. Noticed my language: in touch, in contact, in touch with something. But that’s all it’s doing, it’s just tracking the here-nowness.
Well, that’s really cool! Why do we have this ability? Well, first of all I’m going to propose a way of thinking about this - he doesn’t use this language, but I think it will be helpful! I don’t think it’s in any way inconsistent. This ability to do this is like salience tagging (writes salience tagging on the board). When I touch this (bottle) I am making this here-nowness salient to me. This here-nowness is salient to me. Not the bottle, not even the flat surface because remember I lose all of those particular qualities. All I have is the here-nowness. This is salient to me, and we do this with demonstrative terms like this! Notice the word this is not like the word cat. Cat refers you to a specific thing, meow meow the animal that pretends to love you! Actually I know some cats now that I am actually convinced do actually love me, so I have to amend my usual comments about cats!! But this isn’t like cat! This can go, watch: this (pen), this (bottle) this (wall) this (light switch), OK? It doesn’t refer to a specific thing, it picks out, it does a salience…/ it makes something, it doesn’t make some-thing, it just makes some hereness and nowness!! Sorry for talking about this [like this], but this is how we have to talk ‘salient’ to you!
Now I want you to pick up on something I just said with this. Terms like this and here and now but especially this. These are linguistic terms and they do what is called demonstrative reference. They do not refer to a particular thing they do not refer to the bottle or to the marker or to the wall but this, this, this, this (pointing to these various things) OK? All they do is salience tagging; this and that. Now why is that important? Well, Pylyshyn wants you to understand FINSTING — FINSTING is obviously not a linguistic phenomenon, I’m not speaking in my head when I’m doing this (points to the X’s and O’s) in fact if you try and speak in your head you’re going to mess yourself up — so he is using demonstrative reference as a linguistic analogy for something you enact. So I’m going to try to draw that out by calling it Enactive Demonstrative Reference (writes this on the board), rather than linguistic demonstrative reference. Which I’ve tried to explain to you with this notion of the salience tagging of hereness and nowness.
Why is this so important? WellWell here’s where the analogy can you help me: I need demonstrative reference I need enactive demonstrative reference before I can do any categorisation. Look, if I’m going to categorize things I need to mentally group them together. This is mental grouping: this, this , this (three individual pens), this (the group of the three pens together). That’s what mental grouping is. Mental grouping is to salience tag things and bind them together in salience tagging. So what am I showing you? What I am trying to show you is any categorisation you have depends on Enactive Demonstrative Reference and Enactive Demonstrative Reference is only about salience and here-nowness! You see, all of your concepts are categorical! That whole conceptual, representational, categorical, pictorial… all of that depends on this (categorisation), but this (categorisation) depends on something that is pre-categorical, pre-conceptual. And you say but you’reAnd you say “but you’re using concepts to talk about it!”. Don’t confuse properties of the theory with properties of what the theory is about! Of course I have to use words to talk about it! I have to use words to talk about atoms! That doesn’t mean that atoms are made out of words or dependent on words! I have to use words to talk about anything, and I don’t want properties of my theory and properties of the phenomena of the theories to be confused. I want a theory about, for example, vagueness to itself be clear! I want a theory about a illogicality to itself be logical. I want to theory about irrationality to itself be rational. Do not confuse properties of the theory with properties of the thing being referred to. Yes, I have to use language and concepts to talk about it, but that does not mean that the thing itself is made out of, or dependent on, concepts and categorisation. I’ve given you an argument and I’ve given you empirical evidence towards this claim and they massively, they massively, converge together.
Now notice, this is a fundamental connectedness to reality you’re getting with the FINSTING, with the Enactive Demonstrative Reference, when you’re getting that initial salience tagging, because it’s like the mind being in contact with the world. That’s why Pylyshyn even uses the metaphor of contact! All right so the representational level is not going to give us what we’re looking for. In fact we need to think about ways in which we need to pursue something that is sub-representational. So in Cog-Sci we would call that.../ the representational level is called the semantic level (writes semantic above aspectuality). Because this is the level at which words have meaning or, by analogy, at which representations have representational meaning. So we have to go sub-semantic, we have to go sub-categorical, we have to go sub-conceptual. Now, is that such a bizarre claim? We saw, in Higher States of Consciousness, that people claim to have the most profound sense of meaning and it is precisely ineffable. They reliably, across traditions, across historical contexts, claim that it is not conceptual, it can’t be grasped categorically and they use the language of hereness and nowness to describe it - it’s fully present, it’s like at eternal hereness and nowness. So this is actually not a bizarre claim to consider. Now it’s difficult for us because we habitually identify with — that’s our ego structure m, I would say — we tend to identify with the way in which we are running representations in our mind; inner pictures, inner speech etc...
All right so perhaps we could consider the computational level [as] the level in which we could explain Relevance Realisation because we have found that the semantic level of representations is inadequate. This is often called the syntactic level. Semantics is about how your terms refer to the world. Syntax is about how your various terms have to be coordinated together within some system. So for example you know that there are grammatical rules in English about how you can put certain things together, that’s the syntax. So in computation what were usually doing is we’re thinking about the relationship between our symbols — I don’t mean symbol in the religious sense, I just mean the things that we’re using within, for example, a code or a program or something like that — we’re talking about the relationship between them. Now there’s been a lot of issues around this and I want to point to a core argument by one of the strongest defenders, one of the originators and defenders of the computational theory of mind (writes Fodor on the board). So this is a tradition — you remember, it goes back to Hobbes — of the idea that cognition is computation, we talked about this, the manipulation of an abstract symbolic system, like a generally logical or mathematical symbolic system. The manipulation of that is what it is to think; to think is to do a computation. Now, Fodor has pointed out, and I think these are arguments in many ways analogous to Wittgenstein and you have to remember [that] he’s a defender of the computational theory of mind. He’s considered to be one of the founding figures within cognitive science, so when he criticises it we have to first of all do two things, he died not that long ago, but we have to congratulate him on his honesty as a researcher. The capacity for self criticism is, for me, a demonstrative measure of how good a researcher is. If you’re finding people that are incapable of self criticism in their intellectual pursuits then I would suggest you give them quite a wide berth in how much confidence you place in their work. So the fact that he does that is important and the fact that he launches into that self criticism means he’s not being driven, not being motivated by his own particular theoretical bias. All that being said, what’s the nature of the criticism? Well, the nature of the criticism is you have to make a distinction; ultimately you have to make a distinction between implication and inference.
People sometimes confuse these together (writes implication and inference on the boards). So implication is a logical relationship based on syntactic structures and rules, a logical relationship between propositions. So here’s an abstract symbol: so if I have “A & B” and I know that’s true, I can conclude that “B” is true; I don’t know what B is, see I don’t have any semantic content, it’s purely semantic, but I can derive that. Now, when we try to think about implications, what we have to remember is an inference is when you’re actually using an implication-relation to change your beliefs. And the thing about beliefs is that they have content. So when I’m making an inference I am not just making an implication I am using implication-relations in order to alter belief; changing belief. OK, you say, “well why does that matter?”. Because changing beliefs to us brings up the important issue right away, the important issue right away is what beliefs should I be changing? What beliefs should I be changing? Let me try and show you what I mean: any proposition technically is defined in terms of its logical, syntactic structure by all of its implication-relations, and logicians can get very technical here about whether or not negation and implication are identical blah blah blah I’m just going to speak very broadly here because that’s all I need. So a proposition: it’s logical, it’s computational Identity is defined by all of its implication relations to other propositions. So, for example, part of the identity of this “A & B” is that it implies B; it also implies A and all kinds of things!
Now the issue that we have, and this is a point that was made, also independently, by Cherniak, is the number of implications, logical relations between any proposition and all the other propositions is combinatorially explosive. Combinatorially explosive! You cannot ever make use — and we talked about this, about how you can’t be comprehensively logical — you can’t make use of all of the implications of any proposition, ever; you cannot be completely logical, ever! What you do is, out of all of the implications, you decide which one of the ones you select, which one of the ones are going to be used in an inference. Fodor and Cherniak both independently talk about this as a kind of cognitive commitment: which of the implications are you going to commit to? And this matters to you! It matters to you because commitment is an act that makes use of your precious and limited resources of attention, memory, time, metabolic energy... you cannot afford, you cannot afford to spend them on all possible ones. You cannot even afford to spend them on inferences that are not — and here’s what you knew I was going to say — relevant to the context! Which beliefs do I need to change — and that can mean strengthen by the way — which beliefs do I need to change ‘in this context’?
So notice, what out of all of these (implication relations), what am I doing? I’m choosing — and this is what Cherniak specifically argues, this is his term not mine — what makes, according to Cherniak, somebody rational — we’ll come back to whether or not this is a good definition of rationality but it’s at least what makes you intelligent as a cognitive agent — is that you select out of all the possible implications, the relevant ones because those ones are relevant to the context, because they’re going to affect the beliefs that you’ve already done Relevance Realisation on, as applying to this situation, or representing the situation well. So inference massively presupposes Relevance Realisation.
Now you may think “well, but I can get around that because logic isn’t just implications, it’s the rules governing the implications. And maybe all I need to talk about is the rules!” And then here’s the argument, that comes from Wittgenstein but I think ultimately it goes back to Aristotle, is how rules work, right? And this is an argument that Brown and others have made very, very clear: rules are... obviously they are propositions! [But] they are not just propositions, they’re propositions that — and this is perhaps why you’re considering them — propositions that tell you where to commit your resources. Now the problem with that is that, of course, every rule requires an interpretation, every rule requires a specification of it’s application. Let’s just use a moral rule because they are the easiest for people to have a connection to. I assume that many of you have this rule: “be kind”, which means in a situation I will use inferences to derive actions and changes of belief and those will fit together in a certain way that will result in me achieving kindness towards others. So I have this rule, it tells me which implications to pay attention to, which beliefs I should make salient , etc. Now, what’s the issue about this? Well, think about being kind... what do I mean by this problem of interpretation by specifying the application of the rule? The way I am kind to my son Spencer, what it means to be kind to Spencer, should I use that in the how I’m trying to be kind to my partner Sarah? No! That would be anThat would be an appropriate. It could be condescending. It could be patronising. Now I want to be kind to both of them, in fact I love both of them deeply, but I’m not going to be kind to them in the same way most of the time. Well, what about how I’m kind to a friend? Should I be kind to a friend the way I’m kind to either Spencer or Sarah? That doesn’t seem right either! What about how I’m kind to my students, should that be like I’m kind to a friend? No! How I’m kind to Spencer? No! How I’m kind to Sarah? No! What about how I’m kind to a stranger, should that be like I’m kind to my students? No! How about when I am kind to myself? Should it be like any of those?
So here’s the thing and this is bound up with the fact that we have to always convey more than we can say - you can probably see that! I cannot specify all the conditions of application of the rule in the rule because the rule always has to convey much more than it can say. If I try to specify it in the rule the rule will become unwieldily because it will become combinatorially, explosively large; it will no longer serve. Well, you say, “well what you might do is put in a rule on how to use this rule - a higher order rule!” That’s not going to work because the same problem is going to happen here! And this was Wittgenstein’s point: you can’t ultimately get an explanation of how you follow rules in terms of just the rules. Your ability to follow rules is actually based on something else. Brown calls this, in his book on Rationality in 1988, the Skill of Judgment. Notice what we’ve moved here, we’ve moved out of the propositional language of a rule and we’ve moved into the procedural language of a skill. The skill: knowing how to judge what is relevant, pertinent in this situation. Now again, notice how we can’t even maintain the two things that are supposed to be central to computation: we can’t use inference because it presupposes relevance [and] we can’t use rules because what is this procedural skill of being able to determine what is appropriate or what fits in the context, what fits the people or the situation, what fits the problems or task at hand? Well, that’s the skill of Relevance Realisation.
So we’re seeing that the computational level isn’t going to do it for us. I want to stop here, before we go to this modularity issue, and point out something really interesting. Notice what we got with Fodor and Wittgenstein, and like I said, I think this ultimately goes back to Aristotle... Notice how the propositional — and this was one of Wittgenstein’s famous arguments — ultimately depends on the procedural (writes propositional above procedural with a down arrow between them). One of my favourite quotes from Wittgenstein has to do exactly with this. He said, “even if lions could talk we would not understand them.” Even if they could use all of our words we would not understand them because their skills of what is relevant or important or central to them are very different to ours. He called this “a form of life”. Their form of life, the way they exercise across many contacts the skill of doing judgments of what is relevant, of what is salient and important to them, is fundamentally different from ours because they are cats rather than humans and therefore even if they spoke we would not understand them.
So we see that the propositional actually depends on the procedural (taps this on the board), but notice — and this is really important — if I am exercising the skill, so I’m going to throw this (pen), or do a Martial Art block (demonstrates block), or something [like that], that depends on what’s called Situational Awareness. if I am a good martial artist, I don’t just have my skills and just apply them mechanically — it’s a great thing if you spar with somebody that’s fighting mechanically, because they don’t have situational awareness! Now what is situational awareness? When I’m exercising a skill it depends on my situational awareness. What is situational awareness? Well, you know what it is! We’ve already talked about it! It’s your perspectival knowing, it’s your ability to do salience landscaping [and] it’s ability to foreground, background, formulate the problem… it’s all of that perspectival stuff. So my situational awareness is how my salience landscaping - foregrounding what’s most relevant to the task, [backgrounding], how was it and is it [relevant?] [-]. What’s irrelevant? How it’s adjusting as the situation is changing so that the way I’m applying my skill is more adaptive and more fitted to the situation… (all the while demonstrating martial arts movements). So your procedural knowing depends on your perspectival knowing (writes perspectival below procedural with a downward arrow between them).
Well, you know where I’m going to go with this, right? Your perspectival knowing ultimately depends on how well the agent and arena fit together and generate affordances of action and affordances of intelligibility. If the agent and arena need to be in a conformity relationship, they need to be well fitted together — you’ve seen lots of arguments to this — in order for my salience landscaping to function appropriately. So the perspectival ultimately depends on the participatory (writes participatory below perspectival with a downward arrow between them). Now of course it goes this way, right? (changes the three down arrows to up arrows also.) They affect each other in multiple interactions (draws arrows linking different levels), I was not originally drawing the arrow of causal interaction, I just did that, but what I was trying to draw originally was the arrow of dependence: asymmetric dependence. This depends on this (Propositional and procedural), this depends on this (procedural and perspectival), this ultimately depends on this (perspectival and participatory) (works down through the different levels highlighting dependency). So we are getting a lot about how we should think about Relevance Realisation where we should look for it and notice it’s starting to give us a way of connecting and thinking about the four kinds of knowing.
What about modularity? Well the idea would be something like this, and to be fair this comes up a lot, the idea [that] (drawing) here’s the mind or the brain (draws a circle) and here’s something like the “Central Executive”, or something like that (draws a smaller circle Inside the big circle and labels it Central Executive) — it’s weird we use a business term for an aspect of our cognition, this is used in psychology — and the idea is that the central executive is making all kinds of important decisions. Well, maybe the central executive is responsible for Relevance Realisation and a lot of people — and I know this because I interact with psychologists — they say “well, that’s it! That’s the answer!”. But it’s not an answer, it’s not an answer at all! Because if it’s right, it’s ridiculously homuncular, because what does the central executive have to possess [if] inside the central executive is the capacity for Relevance Realisation? I haven’t explained it! I’ve just pointed to a place and the problem is you shouldn’t.../ So first of all I haven’t explained it, it’s homuncular, and secondly you shouldn’t point to a place! Look, Relevance Realisation can’t be in any one place, it has to simultaneously — you know this we’ve talked about this with how attention works, remember? — you know that you’re always going from feature to gestalt and from gestalt to feature (draws two parallel up-and-down arrows beside each other to demonstrate to this). Attention has to be moving out towards the gestalt and down to the features. Relevance Realisation has to be happening both at the features level and the gestalt level in a highly integrated, interactive fashion. You can’t point to one place and say “that’s where Relevance Realisation is going on” because Relevance Realisation has to be happening at multiple levels of cognition in a simultaneous, self-organizing fashion. That’s why it can lead to insight. And as I said, pointing to any one thing and labelling it is not an explanation. It is a homuncular diversion, that’s all it is!
OK let’s try and draw this all together... what are we learning? What I’m trying to show you, we are already learning something very interesting about meaning making. We are learning what we need, the kinds of properties and processes we need, in order to explain Relevance Realisation. First of all our account of Relevance Realisation — and bear with me on this because there’s an important way in which I am going to modify this — but our account of Relevance Realisation has to be completely internal. Now what do I mean by that? It has to work in terms of goals that are, at least initially, internal to the brain and emerge developmentally from it. Why? Look, any goal in which the brain is representing, or referring to something in the world (draws another circle for the brain and an arrow labeled ‘representing’ shooting way outside) can’t be the place where we can generate an explanation of relevance because in so far as I’m representing a goal to myself, I’ve already got the capacity for Relevance Realisation. The goals that are the originating source of Relevance Realisation have to be internal to the Relevance Realisation process. Now what does that mean? The goals have to be goals that are constitutive. What are constitutive goals? Constitutive goals are goals that a system or process have helped to constitute it for being what it is. And this is especially the case for auto-poetic systems. We’ve talked about this. Living things are not only self-organizing, living things are self-organized because they have the constitutive goal of preserving their own self-organization. To be alive is to have, or maybe even better, to be the goal of preserving the self-organization that is giving rise to you. That is a constitutive goal. Auto-poetic things are self-organized such that they can protect and promote, they are constituted to protect and promote their own self-organization. Which means we should see that there’s going to be a deep connection between your ability to do Relevance Realisation and being an auto-poetic thing because Relevance Realisation ultimately has to work in terms of auto-poetic systems - systems that have goals that are completely internal in the constitutive sense. Now that’s important because that means there’s going to be a deep connection between doing Relevance Realisation and being a living thing.
Next, so when I say internal I mean auto-poetically internal. A theory of Relevance Realisation has to talk in terms of processes that are scale invariant. Relevance Realisation has to act simultaneously at multiple levels — local and global, feature and gestalt — and it has to do it in a self-organizing fashion such that it is capable of insight, self correction. And that means, of course, and that ties in again with, being auto-poetic, that the Relevance Realisation process has to be fundamentally self-organizing in nature.
Now we have a problem here, and it’s a problem that might derail the whole project! It might make it sound like the attempt to give a scientific explanation of Relevance Realisation is impossible! Now notice I’ve been sort of playing between those and treating them as synonymous: a theory of relevance and the theory of Relevance Realisation. That’s ultimately because I’ve been dodging an issue, because I am going to argue [that] you can’t identify them. Because here is what I want to argue, or at least I’m going to state what the argument is going to be and then we’re going to pick it up in the next video: I’m going to argue that we cannot have a scientific theory of relevance, we cannot have a scientific theory of relevance. I’m going to try and argue that that tells us something very deep about the nature of relevance and therefore something deep about the nature of meaning and our attempts to explain, articulate and celebrate our meaning making capacities. But I’m going to ultimately argue that that is no reason for despair because what I’m going to argue is that the fact that we don’t have a theory of relevance doesn’t preclude us from having a theory of Relevance Realisation. In fact it’ll give us a good understanding of what the theory of Relevance Realisation is and that will help us because we will realize, pun intended, that all we ever needed was a theory of relevant realization.
Thank you very much for your time and attention.
- END -
To keep this site running, we are an Amazon Associate where we earn from qualifying purchases
Timothy P. Lillicrap is a Canadian neuroscientist and AI researcher, adjunct professor at University College London, and staff research scientist at Google DeepMind, where he has been involved in the AlphaGo and AlphaZero projects mastering the games of Go, Chess and Shogi.
Sperber and Wilson's book: "Relevance: Communication & Cognition" shown on screen.
Book mentioned - Buy Here
Jerry Alan Fodor was an American philosopher and the author of many crucial works in the fields of philosophy of mind and cognitive science.
John Rogers Searle is an American philosopher. He was Willis S. and Marion Slusser Professor Emeritus of the Philosophy of Mind and Language and Professor of the Graduate School at the University of California, Berkeley.
Zenon Walter Pylyshyn FRSC is a Canadian cognitive scientist and philosopher. He holds degrees in engineering-physics from McGill University and in control systems and experimental psychology, both from the Regina Campus, University of Saskatchewan.
Things and Places - Buy Here
Christopher Cherniak is an American neuroscientist, a member of the University of Maryland Philosophy Department. Cherniak’s research trajectory started in theory of knowledge and led into computational neuroanatomy and genomics.
Book Shown - Minimal Rationality - Buy Here
Rationality - Buy Here
An introduction to one of the major controversies in the modern philosophy of science. The author describes the philosophical problem of rationality and provides a new approach to its solution.