Support the show to get full episodes and join the Discord community.
Nick Enfield is a professor of linguistics at the University of Sydney. In this episode we discuss topics in his most recent book, Language vs. Reality: Why Language Is Good for Lawyers and Bad for Scientists. A central question in the book is what is language for? What’s the function of language. You might be familiar with the debate about whether language evolved for each of us thinking our wonderful human thoughts, or for communicating those thoughts between each other. Nick would be on the communication side of that debate, but if by communication we mean simply the transmission of thoughts or information between people – I have a thought, I send it to you in language, and that thought is now in your head – then Nick wouldn’t take either side of that debate. He argues the function language goes beyond the transmission of information, and instead is primarily an evolved solution for social coordination – coordinating our behaviors and attention. When we use language, we’re creating maps in our heads so we can agree on where to go.
For example, when I say, “This is brain inspired,” I’m pointing you to a place to meet me on a conceptual map, saying, “Get ready, we’re about to have a great time again!” In any case, with those 4 words, “This is brain inspired,” I’m not just transmitting information from my head into your head. I’m providing you with a landmark so you can focus your attention appropriately.
From that premise, that language is about social coordination, we talk about a handful of topics in his book, like the relationship between language and reality, the idea that all language is framing- that is, how we say something influences how to think about it. We discuss how our language changes in different social situations, the role of stories, and of course, how LLMs fit into Nick’s story about language.
- Nick’s website
- Twitter: @njenfield
0:00 – Intro
4:23 – Is learning about language important?
15:43 – Linguistic Anthropology
28:56 – Language and truth
33:57 – How special is language
46:19 – Choice architecture and framing
48:19 – Language for thinking or communication
52:30 – Agency and language
56:51 – Large language models
1:16:18 – Getting language right
1:20:48 – Social relationships and language
Nick 00:00:04 What I’m interested in is humanity. And really the fundamental questions of anthropology are the ones that we’re trying to answer, which is, you know, what is universal about humans? Or what makes us all the same? But unlike all the other species we’re related to, the words are not really meant to capture the specifics of the perception They’re meant to be kind of landmarks that we coordinate around, right? So, um, the, their function really is primarily about social coordination. You know, you’re not particularly free <laugh> when it comes to what you want to express and how you want to express it. You’re very much at the mercy of something of a system, namely in, in our current case, English.
Speaker 2 00:00:57 This is brain inspired. I’m Paul Nick Enfield is a professor of linguistics at the University of Sydney, and he calls himself a linguistic anthropologist, and you’ll learn what that means. Um, in a few minutes. In this episode, we discuss topics in his most recent book, language versus Reality. While language is good for lawyers and bad for scientists, a central question in the book is, what is language for what’s the function of language? And this is a question, um, that people answer in different ways. So you might be familiar with the debate about whether language evolved for each of us to think our own wonderful human thoughts or, uh, evolved for communicating those thoughts and information, uh, between each other. So Nick would be on the communication side of that debate, but if by communication we mean simply the transmission of thoughts or information between people, like, I have a thought, uh, I send it to you in language, and that thought is now in your head, then Nick wouldn’t take, um, either side of the debate.
Speaker 2 00:01:59 He argues that the function of language goes beyond the transmission of information, uh, and instead is primarily an evolved solution for social coordination problems coordinating our behaviors and attention. When we use language, we’re creating maps in our heads so we can agree on where to go. So, for example, when I say this is brain inspired, I’m pointing you to a place to meet me on a conceptual map saying, get ready. We’re about to have a great time. Again, <laugh>, or if you’re my wife’s Aunt Allison, uh, it means get ready to not understand anything I’m talking about. Again. Hi Allison. Hope you and Bruce are well. Um, or if you’re my curmudgeon expert friend, um, I won’t name names here, but, uh, it would mean to get ready to send me that text about what went wrong in the episode. So, in any case, with those four words, this is brain inspired.
Speaker 2 00:02:52 I’m not just transmitting, uh, some information from my head into your head. I’m providing you with a landmark, uh, so you can focus your attention appropriately from that premise. That language is about social coordination. We talk about a handful of topics in his book, uh, like the relationship between language and reality. The idea that, um, all language is framing, that is how we say something influences how to think about it. We discuss how our language changes in different social situations, the role of stories, um, and of course how large language models fit into Nick’s story about language. And I had a ton more to ask Nick. Um, and his excellent book covers a lot more, but we were a little rushed for time. So hopefully I’ll have him back on the podcast soon enough to keep the conversation going. I link to the book and Nick’s information in the show notes at brain inspired.co/podcast/ 166. Um, apologies if my audio isn’t as good as usual. I’ve been in transition for a while, but, um, I’m now in my new place in Pittsburgh, Pennsylvania, and I’m still putting my, uh, new quote unquote studio together, but I should have it back in shape, um, soon enough. So thank you for listening and your patience. Um, and thank you to everyone who supports Brain Inspired through Patreon and through my online course about the intersection of neuroscience and artificial intelligence. You can learn more about firstname.lastname@example.org. And now, this is Nick Enfield.
Paul 00:04:23 Maybe, maybe this is a question, it was gonna be a comment, but I’ll, I’ll make it a question. Um, is it imp? So one of the things that I got from reading your book, uh, is that when we learn language, we don’t learn necessarily how it affects us and how it works, how it affects our cognition and what it’s for. I, is that an important thing for us to learn, uh, in conjunction? And would that solve some of our, our societal problems if we understood more about what your book is teaching us?
Nick 00:04:57 Well, uh, to the second part of the question, I would hope that it would help us. I, I, I, I certainly try to make that case in, in the end of the book, that there, there’s a lot of importance to having good knowledge about language and because it allows us to, I think, be more mindful about the impacts that language has. And I think that’s, that’s kind of one of the key messages in the end of the book. Uh, but going to the first part of your question, you say, when we learn language, um, you know, language, this is one of the things about language. You, you learn it in one sense. You learn it as a child, and you have this sort of incredible capacity, uh, with language to, to use it, to create new expressions, to, um, you know, do all the amazing things that linguists learn about in their kind of undergraduate degree in linguistics.
Nick 00:05:50 Uh, it’s pretty phenomenal in what we do. But the, the, the amazing thing is that kids who learn their first language don’t, so much of that is not conscious. So when you say, when we learn language, um, perhaps you’re talking about when we learn about language or when we learn a second language or something, you know, that is more kind of conscious as, as, uh, as adults. Um, but when we learn our first language and we just gain, uh, linguistic capacity as a, as a functioning human, uh, you know, I think we really do learn a lot of those, uh, aspects of language. Certainly as language producers, we are adept at using language as a tool. We’re adept at using it to influence others. Perhaps some people are more adept than others, but it, you know, my argument and that of many others is that that’s what language really is for.
Nick 00:06:44 So you wouldn’t be learning language if you weren’t also, uh, learning those kind of aspects of languages, pragmatic function. Uh, so yeah, in answer to your question, I think you, I think you need to distinguish between the, the, the, the true learner of a language as in, you know, a two or three year old child who’s acquiring, uh, a language versus the scholar of language who is very often, uh, you know, not paying attention to those much more social interaction, pragmatic features of language, and focusing much more on, uh, let’s say, uh, individual oriented aspects of language, informational aspects of language rather than, uh, aspects related to action.
Paul 00:07:33 But if I, I mean, even learning first language. So I have fairly young kids and they’re adept already at language, and I didn’t tell them, Hey, uh, you should read Nick Enfield’s book and realize that language is not connected to reality in the way that you might think it is. And I’m wondering if that, you know, how you say things matters. I think that we teach our kids that. Um, but, but the, the nature, so just one of the ideas from your book is that, um, our perception is one step abstracted from reality in that we reduce, for instance, the, um, light spectrum, right? Of the, um, the visible light that we can see is only one small blip of this whole spectrum, right? And so our perceptions are a reduced de dimension from what you call brute reality. And then language is a further step abstracted from our perceptions. Um, yeah. And so that there’s this, there’s this kind of two-step, uh, pro connection to brute reality. Um, and I, I think when I learned language, so the, these are things that I didn’t, I don’t think I appreciated at all. You know, maybe even until reading your book, it hadn’t, hadn’t been put that crystal clear to me. And I don’t know, I don’t know how it’s gonna affect my language moving forward. Uh, and I don’t know how it would’ve affected my language from the get-go.
Nick 00:08:56 Yeah. Um, I think I would probably think about things going in a slightly different direction. I mean, I, I feel that things like understanding that brute reality is far, far richer than you can perceive is, is not where you start, right? I mean, that is quite hard to comprehend. And it’s something that in fact, uh, requires language in, at least in part for you to get access to. So over history, humans have made discoveries about the world beyond what we can perceive with our bodies. And, um, you know, we’re able to draw each other’s attention to these things. We’re able to kind of, uh, you know, hack the environment in various ways and discover that there are frequencies we’re not, uh, perceiving and that, and that kind of thing. So getting to understand that brute reality in that fullest sense is, is, is very unusual.
Nick 00:09:49 And, and, uh, we’re not built for that, right? We’re, we’re literally not evolved, uh, to, to know about a whole lot of reality. Um, and so I think that, you know, you could say a starting point for reality for humans is, is what we perceive. And, and, and it’s the band. It’s the spectrums of, you know, light and sound that we can actually, uh, perceive. And over time, of course, we learn to take cues that would help us to know other things about the world. So an example might be, uh, frequencies that you can’t hear. Now, if there’s a, if there’s a sound and a frequency you can’t hear, well, you can’t hear it, right? But, uh, maybe a dog can hear it and they start reacting to it, and then the dog’s reaction is then a sign for you. And, you know, okay, if something’s happening, I don’t know what it is, but you know, the, the dog reacts in this way.
Nick 00:10:43 Um, but even in that case, what, what’s happened is that the dog has kind of transformed the sound into something else that now I can perceive. So there’s that whole sort of translation piece where, uh, you somehow to get a, to get access to a larger reality, it’s still gotta be packaged in a way that you can perceive it. Um, so that’s kind of looking down into brute reality, but then when we kind of look up into the linguistic representation of it, well, that’s another kind of ridiculously dumbed down version of what you know, you actually perceive, right? So, I mean, just take color, um, uh, you can perceive hundreds of thousands or maybe millions of, uh, just discernible differences in, in shades of color as
Paul 00:11:31 Your book, as your bookshelf behind you can attest to.
Nick 00:11:34 Yeah. Uh, but we don’t have labels for all those. And, um, language is very a great deal. It’s very well known literature on how language is different in terms of their color vocabulary. But even the most, you know, even a Dulux paint catalog with, you know, two and a half thousand different, uh, distinctions in pain is still a fraction of what you can actually, uh, perceive. So, um, and, and that’s a, a kind of a crazily over differentiated, uh, set of the, the space for, for very specific kind of economic purposes, but most, for most people in, in life, uh, you get away with a very, very kind of, uh, blunt kind of differentiation of the, of the perceived space. And why, why do we get away with that? Well, because the words are not really meant to capture the specifics of the perception. They’re meant to be kind of landmarks that we coordinate around, right?
Nick 00:12:40 So, um, the, their function really is primarily about social coordination. And I think that’s the thing that’s kind of, there’s this massive kind of, uh, I don’t know, blind spot in some sense, in, in, in the cognitive science of language. I think that, that, that is really, that coordination is not something that you take language and then go and do, uh, with, it’s really the other way around that. So, you, you know, you said at the start of your question, you’ve got young kids and, and you, you know, well, to my mind, young kids aren’t that much concerned about the relation between language and perception. Obviously, they have perceptions and they need to connect language to them in some way or another. But what’s primarily interesting about language for young kids is that they can use it to, uh, manipulate, especially the other agents in their vicinity.
Nick 00:13:36 So I’m sure that your kids are very good at manipulating you and, uh, no, no, no. So you say, um, you know, they’re not sure how old they’re, but, you know, the, from a very young age, um, not just using language of course, uh, you know, so, uh, gestures, uh, very powerful at a, at a young age. But as soon as they get on to the idea that, um, words create responses in other people, uh, quite specific ones, then, you know, that’s what becomes interesting about a, about a word. And, and I, you know, I think that it’s their embeddedness in sequences of action, their embeddedness in, uh, people’s courses of behavior that makes them interesting. So, take a simple example. I don’t know, something like, um, uh, a verb like open, or I guess an adjective too. So, to, to be open or to open something, you know, you can argue about what exactly that word means, and what is the physics of something being open or closed, and when would you, you know, you could study the reality of things that are open or closed and, and, and argue about what the, what the concept is.
Nick 00:14:45 But if you look at how that word gets used with young children, well, it’s gonna be about things like we open the box and, you know, then an activity begins. That’s the toy box, for example. We pull out the toys and we play, and then it’s over, and we have to close it, and kids will associate open with the sort of beginning of an activity and close with the, the end of an activity. And you have these associations that are not really about kind of measuring the physical world, but they’re about being embedded in a, in a sequence of activity and action. And that becomes, uh, the initial, meaning the initial association, the most important kind of association is how words link to courses of action and particularly interaction.
Paul 00:15:32 So it, okay. Well, this opens up, uh, many things that I wanted to ask you about. Um, maybe though, I would like to back up, and I, I should have done this in the beginning because y your title or, or the title of your vocation, uh, the name that you give to your vocation is that you’re a ling linguistic anthropologist. And, um, one of the things I enjoyed from your book, um, is, you know, reading about all the anecdotes, uh, from your travels, and it just struck me that you have, through your work, lived a rich life interacting with so many different societies and languages. Um, and so I guess my question is, have you lived, uh, uh, do you feel like you have lived a rich life through your work? And then what is a linguistic anthropologist? So, so that I can understand it myself,
Nick 00:16:24 <laugh>? Well, I’m still living, I’m still living. So, uh, hopefully many more decades thus
Paul 00:16:29 Far, of course,
Nick 00:16:30 <laugh>. Yeah. Um, yes, absolutely. I’m incredibly fortunate and incredibly grateful that, you know, I’ve had an amazing, um, a number of experiences in cultures very unlike my own and, and places that, that aren’t, uh, like the place where I grew up. And, um, that’s I think probably something that got me into what I do is, is a fascination with travel and, and, and with diversity, what’s a linguistic anthropologist is, um, really I identify that way because what I’m interested in is humanity. And really the fundamental questions of anthropology are the ones that, that we’re trying to answer, which is, you know, what is universal about humans? Or what makes us all the same, but unlike all the other species we’re related to? That’s sort of one question. And the other question is, what are the constraints on the diversity within our species? So, you know, we have these incredible differences. We have these thousands of languages, we have these different livelihoods and, and value systems, and what is the, what are the possibilities for diversity? So that’s the kind of fundamental question of anthropology, which has been around for a long time. Uh, and it, it, it’s not so much you do still hear this term, but there’s a, there’s a, there’s a notion of anthropology as having four fields and linguistics, uh, was originally conceived as one of these fields. Oh,
Paul 00:18:03 Okay.
Nick 00:18:04 Yeah. So, um, when people use the term linguistic anthropology, sometimes they mean, you know, it’s that field of anthropology that is concerned with looking at, at that universal’s diversity and language these days. Of course, that’s kind of captured within it in many fields of, of linguistics itself. And, um, anthropology’s kind of developed in, uh, the four fields’. Conception is not necessarily used by, by everyone, but that’s how I view it, is that essentially I’m looking at language as a way of answering the basic questions of, of anthropology, which have to do with human universals and diversity.
Paul 00:18:40 Ah, okay. All right. Thank you for that clarification. Um, and I, I wish you many, uh, happy fulfilling years ahead, by the way, uh, and continue to, yeah.
Nick 00:18:50 Uh, look, I mean, get, you know, the, the, the work that we’ve done, uh, in linguistic anthropology, I mean, this has been going on for a long time, uh, way before me, of course, but, you know, it does, uh, coming back to your rich life question, I mean, it’s doing field work and getting out into other cultures and trying to understand them and participate with them is just so enlightening in some sense, right? It, it, not only are you learning something, but what you’re doing is changing your kind of calibration. You’re recalibrating yourself in some way. And, um, you know, that’s obviously important from a scientific research point of view, but also it’s, it’s, it’s good for you. Uh, you see what your own presuppositions are, hopefully. And, um, you know, that that helps to draw your attention to the things you’re taking for granted. So I think, you know, uh, in terms of scientific approach, that’s important, but I, it’s also, it’s also good, uh, as a general principle. Hmm.
Paul 00:20:00 I was gonna ask you this later, but I’ll ask you now. Um, and I know that we haven’t gotten into many details about what, what your ideas and work actually say, but, um, you know, has your work affected how you use language and interpret language, like kind of on a daily basis? Do, do you have an on and off switch, or are you constantly thinking about the framing of what’s being said and the influence and the social coordination dynamics and stuff? Or are you just sometimes talking and listening?
Nick 00:20:29 Well, I’m talking a lot, most of the time. I mean, it’s hard to introspect about these kinds of things. And I, you know, one of the things we, we all learn early on in, in linguistics is that people are quite poor at, at introspection, and they, they believe certain things about how they speak, that that empirically just a false, you know, they say, oh, I would never say that. And then in the next sentence, they say that thing, or, you know, I, I only seldom use that word, but then you record them for a day and they use it all the time. So, you know, there are, we’re quite poor at that sort of introspection. Um, but, but I’ll respect anyway, I think that for me, a lot of the time, of course, I’m just talking and I’m not kind of thinking carefully about this stuff.
Nick 00:21:15 And, and, and one of the reasons, of course, is that who has time? I mean, you, your processing language so quickly, so much has to happen in such a small amount of time when you are, when you’re just having a conversation that you, I mean, you literally don’t have the bandwidth to, to kind of worry about those things. And when we do all this stuff in a, in a, in a fairly unconscious way, but I do think that it enters into more deliberate use of language. So, you know, you’re giving a talk, you’re doing something a bit more formal, you’re in a meeting, you know, when you have some, you know, there’s some issue with your kids or whatever you, you, when there’s some kind of, um, more heightened importance in a way to the situation, then yeah, I think I do become much more, uh, aware, um, I pay more attention to the framings, uh, not just as a producer, but also as a, as a comprehender of language.
Nick 00:22:11 And, and I think that’s helpful. And I think in this day and age, it is especially important to be very critical in a way of how framing is being done when you are on the receiving end, because, you know, now it’s just so, um, in terms of the kind of the media that we’re being exposed to and so on, you, you, you really have to be a, a, I’m not sure if you have to be more careful than, than we used to be, but just, just quantitatively the, the sheer amount of stuff we’re being exposed to, um, I think is, is is means that we do need to be more mindful. Uh, so coming to your question, I mean, I, I think probably the answer is yes, I’m becoming more mindful of how language is getting used, and that’s what I advocate in the, in the book. So hopefully I’m, I’m going some way to practicing what I preach.
Paul 00:23:09 Yeah. Yeah. I mean, you mentioned media and, um, I see if this, see how you feel about this. I’m not sure about the state of your news outlets, um, in Australia, but you know, here it’s all about fake news. And I’ve decided that no news is real news. It’s all framed, and it’s all just influence. And my parents’ generation and the generation before them, it seemed like it was different, because I don’t know if the framing was different, um, or the nature of influence in media was different, but it’s, it’s like you could kind of take some bit of news at face value, and, and I’m wondering, here’s the question. I’m wondering if, like, my mom, um, still believes the news because she was raised in an environment in which you could believe the news. And so, you know, whatever Fox News or whatever is telling her, uh, she’ll just take it kind of face value, even though it’s an opinion show or something like that. And, and have the news outlets just learned to frame everything in a very specific way, or, you know, do I, do you think that that’s right about my parents’ generation and, and their parents’ generation, that they could kind of trust the news more? And now that’s why these days we we’re in this predicament where they’re being fed false information or heavily, heavily framed information,
Nick 00:24:27 Uh, I think it’s easy to overstate that point. I mean, I think that, you know, we have this tendency to say, well, those were the good old days, and now kids these days, and that kind of thing, you know? And I think that certainly not the case, that if you went back 50 years or something, that suddenly everything would be hunky dory and the, the, the newspapers would be full of true facts and all of that. So one thing that’s clear about newspapers, you know, 50 years ago, is that they, that, that they’re very selective. You only, uh, read about certain things and not other things. And that’s already a big kind of difference. And, and, and that is less the case now because we have so much more material. So, you know, as a, as a consumer, you have much more choice. You can get much more information than you used to.
Nick 00:25:11 So there are clear ways in which, you know, the, the information ecology is, is better and healthier because you have access to more information. Um, but because of that, um, scaling up, uh, of, of material and also across time, so now just endless 24 hour, you know, news, that means that the audience capture, you can still, you know, you don’t have to capture the whole, uh, population, you know, you’re just going for a particular group. And, and this leads to the kind of classic polarization. So, I mean, I think the thing to really not forget about news is that it’s, it’s, it’s a money-making industry. It’s, it’s just a business, right? So information is not really the primary concern as a, in terms of like getting at some version of the truth. Uh, that sounds kind of cynical, but I, I think it’s true. And, uh, I think it’s something always worth worrying about that, and being critical about is going back to what I was saying earlier, that you are always consuming something that has been produced that way for a reason.
Nick 00:26:24 Um, and, and that’s the case just when you’re having a regular conversation, right? Different dynamics, of course, and different things are at stake. Um, but, but I think that there are ways in which news and information is worse now, and ways in which it’s better now. But one of the things that’s really highlighted today is that the importance of being our own individual gatekeeper, uh, people, you know, I was asked some time ago on a kind of panel about fake news and, and post truth and stuff, well, who should be the gatekeeper of, of, of, uh, facts, you know, we need, uh, fact checking and all this kind of thing. And I think, you know, the, the days of excitement about fact checking from a few years ago has kind of waned because, uh, partly it’s so hard to do, um, uh, if you can even do it at all.
Nick 00:27:22 But I think it’s because it’s, it’s looking at things from the wrong direction. So it’s really about, uh, critical thinking. It’s a kind of, um, it’s, you know, it’s about, it’s cognitive literacy I like to call it, which is to say, um, you know, we don’t need as consumers of news or any language, um, we don’t need a filter before it hits us as if, if we have our own filter, uh, that we’re able to apply in, in, in a, in a kind of mindful way. So if you went back to the news that your, your mom, your mom was reading, um, you know, as a, as a kid or a young person or whatever, um, it, it would have different properties, but you would certainly need to read it mindfully if you wanted to get at what was actually going on. So the, the news of that time might have been quite balanced, let’s say, or if, you know, re relatively objective, um, for that audience.
Nick 00:28:21 Uh, but just look at, well, you know, if it’s reporting some kind of foreign policy or some kind of, um, event that’s happening overseas, what are we getting any perspective from the people on the ground over there? Um, seems unlikely. Um, are we hearing about, um, things that are happening between groups over there that don’t, that don’t sort of concern us? It seems unlikely. So there’s a whole lot of things that have been filtered out and, and, and there’s just as much need then for reading news in a discerning way as there is now. Hmm.
Paul 00:28:56 Well, yeah, I don’t wanna, I don’t want this to devolve into a political, uh, <laugh> episode, obviously, but, um, but thinking about, so, so when, you know, you argue that the main function of language is for social, social coordination of, um, social problems, essentially. And we, we already talked about the two degrees of separation, you know, between language and reality. Um, does, does that mean that language cannot describe truth? And now I’ll back up and, you know, obviously when I write a science paper, I’m telling a story and I’m using language to do so. Um, should I be concerned that there’s just no way that I can, uh, describe something the way that it actually is? Am I des, am I trying to describe truth or am I telling a story to influence the reader that, uh, it is true? You know, what, how, how do I, how should I think about language versus truth?
Nick 00:29:55 Well, you can do both those things, you know, hopefully, uh, the thing that you’re saying is true, and you can convey it to a person while through, through the tool of influencing them, uh, you know, so that’s what what you want to do, is to get the kind of influencing function of language to align with, uh, some sort of fidelity around reality. First thing I’d say is there’s a, there’s a distinction between reality and truth. Um, I think so, you know, reality is the stuff we come up against, the thing that makes us die, et cetera. Um, truth is really a property of, of assertions that you make in language. So, you know, is that true or not? Is is not a question about the world. It’s a question about your claim about the world. And so really, it’s, it’s about language. The question of truth is really about language.
Nick 00:30:52 And, and the claims that we make are very vague, and they’re very underspecified, and that’s by design. So, you know, if you want to coordinate with someone around some statement, uh, then you don’t want to go to too much trouble. You just want to make a statement that’s good enough for the purpose. So it’s a classic kind of, you know, MinMax trade off kind of thing. You, you, you want to go to as little trouble as you can in formulating your utterance, um, in order to achieve the goal. So, you know, you might say something like, uh, there’s a dog in the backyard. Um, you know, and whatever the situation, you can imagine why I might say that to you, maybe I’m, uh, checking, is that your dog or is that normal? Is it like the neighbor’s dog? Or what are we gonna do about this?
Nick 00:31:44 Or whatever. Um, but I use the word dog, and that’s kind of this category term for really great variety of beasts, right? Um, when I say there’s a dog in the backyard, I don’t tell you anything about this dog. It’s color, it’s shape, it’s breed, how it’s acting, all these things. Um, because for the purpose of that statement and what I might be trying to kind of what re response I might be trying to get outta you, those details are irrelevant. So I have just the level of specificity around reality that I, uh, that I need for this, for this, uh, function that I’m trying to serve right now. Or, or, you know, I’m, we’re at the, the, the, the table and I say, can you pass me a spoon? Um, well, I don’t mind plastic spoon, metal, spoon, whatever, just for this, for the purpose of what I want you to you to do.
Nick 00:32:35 Now, this word should be good enough. So in relation to your question, what specific spoon it is and what its real measurable qualities are, is not at issue. Hmm. For why I’ve selected this word or that specific dog, you know, it’s measurable, real qualities, uh, are not what’s at issue. Even though I could, they might become, they might come, if you ask me always, is it, is it a black dog? Then it matters because okay, maybe the next door’s dog is black and it comes through the fence sometimes, or whatever the case may be. Uh, so there’s a, a, a, it’s really important with language that we, you know, the concern is not just does this expression or this choice of words bear a relation of being true or not true in relation to what I’m talking about. It’s really what is it being used for? What is the response? It’s being used to elicit what task are we trying to accomplish by using this, uh, term? And that might be a joint task. We might be cooperating towards some end, or it might be some sort of manipulative task on my part, but either way it requires coordination between, uh, the, the producer and the comprehender of that bit of language. Does that answer your question? Yeah.
Paul 00:33:55 Uh, yeah. I’m also, um, <laugh>, so I’ve, I’ve been I guess self-critical or critical of lately on the podcast in a kind of a devil’s advocate sort of approach. I’ve been critical of language as, uh, having the, you know, sitting at the, um, in the sitting at the top of our cognitive capacities, right? We’re so special because we have language. And I’ve been, uh, sort of testing, you know, that hypothesis somewhat, um, informally, right? Saying, you know, cuz I feel dumb when I think in words. Um, and, and, you know, thinking about the connection of language to our actual experience, which is, uh, of a much higher, um, uh, granularity or lo well much finer scale granularity, right? And you take like a concept or a word and it reduces it, um, away again, one more step from away from brute reality. Uh, and so my question is, do you, how do you view language in terms of our, uh, cognitive capacities and, you know, how, where is it in the hierarchy of our cognitive capacity?
Paul 00:35:05 And a related question, and maybe this can kind of, um, funnel you into that, the answer to that is, you know, there are people who literally talk out loud to themselves all the time, or, uh, can, you know, can’t seem to function without speaking. And there’s this term called process blocking that if you and I are at a party and I’m just yammering on like this as I am, you can’t even think straight because I’m interfering with your process because I’m influencing you. Right? Um, so can people spend too much time in language land, uh, in their, uh, daily lives or, you know, is if if someone’s just talks so much, does that mean that they’re higher cognitively, et cetera?
Nick 00:35:45 Yeah, I, that’s a interesting question. Um, I don’t know if I, if I have any kind of opinion on that. You know, people who talk a lot and people who don’t talk as much, it, it, that puzzles me a lot. And I, you know, it’s not something I’ve studied, but it’s something I’ve thought about. We we’re talking, I was talking with colleagues yesterday about students in, uh, you know, university classrooms that some students talk all the time, and then other students are dead quiet. And often there’s an assumption that, you know, a student who’s sort of always piping up is kind of a, a really good student and does well, but, you know, you get this, some students who does seem to say a thing, uh, and then they just ace the, the assignments and exams mm-hmm. <affirmative>. Um, so, you know, I think that we don’t want to over associate talking with kind of some capacity to think and reason. And I think this is one of the issues that I keep coming back to in my, in my work and certainly in this book. And that is that we we’re, we’re always, when we’re talking about language in a kind of scientific context, we, we always seem to be coming back to thinking and reasoning and concepts and cognition and individual, internal, uh, ideas, which is
Paul 00:37:12 The exact opposite of the message in your book.
Nick 00:37:14 Exactly. So, you know, that, that really, you can have all of those things, images and ideas and reasoning and, and, uh, categorization of a kind, um, without having language at all, right? I mean, you can be a creature that doesn’t have language and you can reason about the world and, and act in, in accordance with it and so on. And we are the creatures of that kind, and we have all that internal cognitive life too. Uh, but having language makes a big difference, I think, well, I’m gonna say in two ways. I mean, one is because of all of this public stuff that we do with language, which is about manipulating others in very specific ways, putting descriptions of things on record, and by which, you know, any description is a very specific framing of something that’s going on. Um, you know, highlighting certain things, deemphasizing certain things, um, and, and making claims right about what’s in front of us.
Nick 00:38:14 Uh, you know, is this a dog? Is this a mutt? Is this a pet? You know, I can, I can frame this, uh, thing in, in certain kinds of ways for social interactional purposes, not because the reality kind of gets me to do that. So once we build up these labels, these words that have rich kind of ideas, or maybe not very rich, um, sometimes pretty stripped back, I guess I, I’d wanna say. But we build up these labels, we build up these categories in, in language, and of course then that’s gonna have some effect back on the kind of inner conceptual, uh, life. So I recently wrote something where I tried to kind of sharpen a distinction, which I feel is really not often made and surprisingly not often made. And that’s simply the di distinction between non-linguistic concepts and linguistic concepts. Hmm. So, you know, uh, if you look into the concepts, literature, literature about, um, yeah, I mean, just pick up a book that has the word concepts in the title.
Nick 00:39:20 And, uh, you’ll see people are talking about things like dog and cat and doorknob and walk and, you know, uh, on, and all these things that happen to have labels Yeah. In, in a language just so happens they have labels in a language. And, and my argument, um, is that, you know, that really conflates two very different kinds of, uh, mental entity, right? So, uh, one’s famous paper by, uh, Frager, uh, that’s often cited for distinguishing between sense and reference, where reference is, what a word that’s something in the world that a word points to and sense is sort of the meaning of the word. But in that paper, he also pointed to what he called idea. So, uh, idea is really I think what you were referring to just before, which is, you know, you are completely rich in a experience of something. So let’s say I say the word dog, and you owned a dog as a kid, and all your experiences of dogs and you know everything about them, all your sensory perceptions and your grounded cognition of dogs and everything else.
Nick 00:40:34 Um, but that doesn’t come into issue when we communicate using the word dog, right? Uh, so that your idea is this very individualized, very rich, personal kind of <laugh>, uh, mental thing. And the sense of the word is this other kind of concept. So I talk about I concepts and l concepts in this, in this, uh, recent thing I’ve written. And that’s to point out that, you know, if you want to talk about linguistic concepts specifically, it’s a very different beast. It’s something that is built through interaction for interaction. It’s something that’s gonna necessarily be, uh, stripped back because it, it only has to be good enough for communication, not for making decisions about how I deal with this creature when I encounter it in a dark alley or whatever, you know, individual decision making processes have to take place. These things are created by and created four, uh, social coordination.
Paul 00:41:37 Is, is this, are you referring to the self-generating, uh, idea of a concept paper?
Nick 00:41:43 Um, well, no, that’s another recent paper that is, um, related to this stuff. So I, um, this is a paper that’s not, um, published yet that I’ve just recently drafted about our, our i concepts and L concepts. But the, the, the other paper that you are referring to was, was, was recently, uh, published in the transactions.
Paul 00:42:03 Yeah. Cuz I, well, I wanted to ask you about that. I, I like that your conception of a concept as a self-generating thing, which is, and, and, and that you have two different axes and we don’t need to get, you know, super technical about it, but can you just talk about your, I’ll call it a processional or process-based view of what a concept is? I I think that use the term dynamic, uh, in, in the paper to describe how you view linguistic concepts.
Nick 00:42:31 Yeah. So in that paper, uh, a as you just said, I, I invoke the term choice architecture. Um, you know, there’s this literature on nudge theory, uh, and on the idea that, but, but also a sort of large literature on design and how, you know, the way that you design a door handle, for example, will determine whether people reach for it with a hand shape like that or with a hand shape like that. Um, so the environment sort of speaks to you in some sense. That’s a classic kind of, uh, old insight about how we interact with the environment. Um, so if you’re a designer, then what you’re doing is, you know, let’s say you’re designing, uh, doors, or this is a famous Don Norman kind of, um, work on, on, on how doors get designed. If you’re designer, if you’re a designer and you’re designing a door, your specific choice of how that door will open, will, should communicate to people what they’re supposed to do with their body.
Nick 00:43:33 Like they press it and push through, or they grab a door knob, you know, so that the, the door communicates with you and, and that determines your choices. So it’s, it’s a, it has what we call choice architecture. And so there’s a, an actual interaction or as a dynamic relation between how this static object is structured, uh, physically the door and what you do when you encounter it and wanna functionally engage with it, right? You wanna open the door, well, you have to sort of make choices of how you’re gonna do that with your body. So it’s a, it’s a, it’s a sort of stimulus response, uh, type of axis. Um, and that’s been stripped out of how we think about language, um, for a long time, famously going back to kind of anti behaviorism and so on. Um, the problem is it was sort of thrown out with the bath water and the real, what you really want to do is combine a view of language that is, there’s one axis, which is, uh, uh, you know, uh, some bit of language as a stimulus that gives rise to a response and someone else.
Nick 00:44:38 Um, and another axis, which is really the con the cognitive axis, which is, you know, some bit of language maps onto some mental representations that I might have. So, uh, yeah, let’s pray as far as we want to get in terms of the technical side. But, but it, it becomes a dynamic process because anytime you encounter a bit of language, it’s like a door that you have to decide, you know, how do I shape my hand in order to open it? You, you get a bit of language directed to you and you make decisions. How am I gonna respond to this? Um, how am I gonna adjust my beliefs? Or whatever it may be, whether it’s interaction or you’re just reading something, it’s affecting you and it’s you’re updating, you know, who and what you are and what you’re gonna do next kind of thing.
Nick 00:45:23 Uh, and that very kind of action based aspect of, of, of how language works and what it’s for is, uh, I think underappreciated in, in, in how we think about the function of concepts. So from that, that is l concepts, a linguistic concepts from that point of view, a linguistic concept is something that is designed partly just through, you know, emergently in a population, but also locally. I’m designing my utterances to affect you in a particular way. And in that sense, it’s a piece of choice architecture. It’s something that is directing you, nudging you in a certain way, whatever I’m trying to say. I could have said it in these other ways, but I’m, I’m not, I’m saying it this way. And hopefully that will set you in the direction that is the desired one, either desired for me or desired for us, um, uh, whatever the context happens to be.
Paul 00:46:19 So all language is choice, architecture and framing whether we are intending it that way or not.
Nick 00:46:27 Yeah, I think that’s true. Uh, I think that it’s just something you have to live with. Uh, and one reason I think that helps to make clear that that’s the case is that, you know, we’re never talking about language in the abstract where when we’re talking about language and concepts, we’re always gravitating to particular languages. Some language like English or Japanese or French or, you know, six or 7,000 languages that are spoken in the world today, and who knows how many hundreds of thousands have been spoken through history. Um, you know, each one of them is going to present a very different array of available concepts, um, which don’t exhaust the possibility space by any stretch of the imagination. So, you know, just the, the, the, the, the resources that you’re drawing on when you’re using language are this small subset, uh, of, of all the possibilities.
Nick 00:47:28 And that should indicate to you, um, that it’s a very biased system, right? So that you even, I I, I can take words and I can combine them in new ways and say new things that you haven’t heard before, but I still am limited to the resources that my language gives me, and I can’t reinvent my language in any serious way on the fly. So what that suggests is that, um, you know, you’re not particularly free <laugh> when it comes to what you want to express and how you want to express it. Um, I mean, you’re a bit free, but you’re very much at the mercy of something of a system, namely in, in our current case, English that has been developed through history within a particular cultural, uh, context that, that we don’t have any control over.
Paul 00:48:18 Hmm. So, um, I, I realize that we’re kind of all over the place, so I apologize, but going back to this idea of language as a, um, coordination solving, um, entity, cognitive function, function, um, so in, I guess the cognitive sciences, there’s this kind of classical debate, whether language is for thinking or for communication, and your ideas, uh, I guess would lean more toward the communication side, but it really goes beyond that because it’s about agency and coordinating in the world. Do I, do I have that right? Is that the right way to look at it? Yeah,
Nick 00:48:56 Well, absolutely. So, uh, I mean, there’s a few things to say here. Um, one question is like, what do we, what do we use it for? Uh, just empirically, like, what’s, what do we use it for? So you could go out and somehow count instances of, you know, people communicating with language versus people using it to think if you had some way of kind of doing that, um, I think that, you know, one answer would be, well, that’s irrelevant. You know, the, the modern in this day and age, we’ve realized this wonderful epiphenomenon, uh, of, of linguistic thinking, which is that now we can talk to each other and let’s do that a lot. Um, so I don’t think that would change people’s minds about kind of what’s fundamentally, you know, different or interesting, um, about language. But what I, uh, would say in response to that is, um, it’s not just what a language gives rise to what it can be used to create, like, you know, a new thought or a decision or, uh, an action on behalf of someone else, but, but, but just as important as where did it, where does that particular bit of language come from?
Nick 00:50:06 How was it, how was it brought about? And uh, this brings me back to the point I was just making about how, whenever we talk about concepts, where naming particular words in particular languages, how did those things come about? Well, they came about through social interactions, and through those social interactions are the only thing that cause the words that we’re talking about to circulate within populations. So that, you asked me about that paper, about self-generating choice architectures, and we talked about the choice arch architecture part, but not the self-generating part mm-hmm. <affirmative> and this dynamic kind of cycle of, of, of, of, of bits of language being used to give Verizon to responses, um, in, in social interaction that itself causes bits of language to circulate within a population. So, uh, I like to quote, quote this, uh, line by Daniel Deni when he is talking about kind of circulation of, in diffusion of innovations in populations of new tech.
Nick 00:51:13 And, you know, he says, okay, well, a, a, a wagon, uh, wheel is this wonderful invention that can, you know, help you transport stuff from place to place. Uh, but what it also does is it transports the wonderful idea of a, of a wagon wheel mm-hmm. <affirmative>, um, with it, you know, so when you see that, uh, device being used, you go, ah, that’s a good idea. I’ll do that too. Um, and it’s exactly the same with language. So you, you can’t build co uh, linguistic concepts, which are the kind of concepts we’re mostly talking about in, in, in cognitive science. You can’t build that kind of concept without getting your data from interaction. Your data, uh, that you use to build up a concept of, like dog, for example, or any other of the thousands of concepts we have, um, comes, is shaped by the evidence you get from how people use words and how they respond to how words get used. And, um, most of the concepts are different and most of the languages that exist. So, uh, that to me is a very powerful point about, uh, the primacy of communication and the pri primacy of social interaction in thinking causally about linguistic concepts.
Paul 00:52:31 So agency is central to your, uh, conception of what language is the function of language.
Nick 00:52:41 That’s right. Yeah, it is. Um, and that, you know, you can take that to simply mean we do stuff with language. Um, and that’s really, you know, what a big piece of what I wanna say is that, you know, you, you use language to create effects in the world, and then your effects become, you know, you can track the success or night of those effects, and you can refine language on that basis and become, uh, better at it. I think a lot of what we talk about when we talk about cognition and kind of internal processes, um, is very often about action anyway. So, um, I mean, a couple of examples, um, classic kind of William James definition of cognition saying, you know, you have higher cognition to the extent that you can consider different means to one, to an end mm-hmm. <affirmative>, uh, that you can, you know, um, r Romeo seeks Juliet and, uh, the door is locked.
Nick 00:53:51 So <laugh>, he’s, he has higher cognition and he can find different ways to get to her by, you know, using his, his, uh, his intellect. Um, so we talk about that as being something kind of mental, but it’s necessarily mapped onto action. He’s thinking about these different, uh, possibilities precisely because he wants to make a decision and undertake an action towards a goal in the world. Um, so that’s kind of one example to me. You know, cognition generally is, uh, is equally all about action, right? I mean, that’s what it drives. Um, and then actions and their outcomes are what provide evidence that allows you to kind of update your concepts and your kind of, you know, re reweight aspects of kind of how your reasoning works and so on. Another example related more closely to language is, is the, the, the writing of Benjamin Lee Wharf, who’s, uh, always cited in connection to, to linguistic determinism and linguistic relativity.
Nick 00:54:52 And everyone talks about his work as being about language and thought, uh, language influences, you know, how we think. But in, in probably his most famous paper, it’s, uh, in the title of it, it’s something, uh, I can’t remember the exact title, but it has thought and behavior in the title. And he’s j just as much talking about thought as he is about behavior. He’s talking about actions people took, uh, given certain linguistic inputs that led to them, you know, reasoning in certain ways that turned out to be not very smart. So these were examples from people who started fires in, uh, uh, because they misunderstood something about the situation. Um, so I don’t know, there’s a, a, there’s a, a substance called spun limestone, and, uh, people reported thinking, oh, well that’s, um, it’s just some kind of stone, so it’s not gonna burn.
Nick 00:55:49 So I’ll place this, you know, uh, very hot thing on it. And then a fire starts and they surprise, and I think this is what’s overlooked a lot, is that language played a role there, not just in thought, but in action, right? So a person took a decision and carried it out. Um, uh, uh, so that’s kind of one piece of it. And the second piece of it is that they, that the people in question used language in a different kind of action, which is explaining what they did after the fact. So when they meet the insurance inspector who’s coming around saying, how did this fire start? Uh, you know, who knows what, uh, decision process they went through at the time, but they’re certainly using language in a particular way to, to narrate what they think happened or to explain or to justify, uh, the actions that they took and, and they reach for language to build such an explanation.
Paul 00:56:49 So I’m going to, uh, I’m gonna bring us back. I wanna be sure to bring us back to the ideas in your book here. Maybe we will, um, get to some via my next line of questioning. Um, I’ve been the past handful of episodes, we’ve been talking about large language models a lot, and this seems like a natural time to bring them up since agency is central, uh, to your concept of language and language as coordinating, um, solutions to social problems. What does this mean then? I’m sure you’ve thought a lot about large language models. What, what does your approach and, and ideas mean about how to think about what large language models are doing?
Nick 00:57:32 Yeah, I have been thinking a lot about them. Uh, and so has everyone. So, you know, yeah, there’s a kind of, uh, sort of partly digested thoughts here. I mean, they’re obviously super interesting, the large language models that everyone is engaging with right now and kind of using, uh, one thing that, uh, I would always want to kind of keep in mind when thinking about something like a a, an ai, um, some kind of artificial agent is that we, uh, that we as interpreters of the language that such a model of producers are doing a lot of the work. Uh, and so, you know, you, you, you might say, well, uh, you know, this is just, um, an algorithm and it’s spitting out something, you know, it’s sort of in some stuff went in and now some stuff comes out and you are attributing some kind of mind to that, and that’s the wrong way to think about it.
Nick 00:58:43 Um, I, I think what’s really important is to, is to get away from that focus on the model as an individual mind or some analog of an individual mind. Uh, if you think about this in that more dynamic kind of processual way, every time that, uh, a a, a large language model spits out a a piece of output, um, it’s not just this disembodied thing, it’s, it’s a, it’s something that you then process. And so there’s this relation between the, uh, the language that the model produces and your interpretation of it. That, that, that’s, that’s really what we’re interested in here. So, you know, you as a user, as an interpreter of this output are, are putting a whole lot of the meaning in, uh, just in, in identical way to what’s happening when you’re just having a conversation with a real human being, right?
Nick 00:59:37 So you’re it, you’re never getting a direct download of what the person thinks or means or wants to say. What’s happening is that they are giving you some set of clues, and then at the other end, you are making all sorts of inferences and filling in intentions and filling in kind of, uh, you know, information that you, that actually wasn’t expressed, but that you think you can infer based on what they said. So it’s a kind of classic kind of differential model of language understanding. And we apply that to, to, to these models and it feels very natural, um, and it works really well. So I think, you know, there’s a, there’s a great sort of functionality to it. The big difference, of course, is that, uh, it’s really about the social relational element to how you relate to that particular agent as an agent.
Nick 01:00:33 Okay. So, um, foundationally, one of the big things I argue is that, and, and in, in, in other work as well, um, is that when you are using language and you’re using it to, to carry out social actions, there’s this big piece, uh, to the process which we, I typically refer to as accountability. Um, and so this is something that is big in the kind of micro sociology of language. Um, I published a book recently with Jack Sid called Consequences of Language, where we focus on this kind of accountability, um, piece. And what this is about is looking at how language forges social in relationships and how it sets up kind of, let’s say, um, well, I’d said accountability, uh, reciprocity of sort of roles and so on between people and language is what allows you to really set up your, your personal relationships and maintain them.
Nick 01:01:31 Now, when it comes to an ai, um, you know, what’s your relationship with that thing? Well, you don’t have that kind of accountability, um, uh, or it doesn’t have that kind of accountability in some sense. Um, you, it’s a very different dynamic, and I think that we’re just a beginning of learning how to actually interact with that kind of, uh, with that kind of an entity. And I think we’re, we’re, uh, probably making all sorts of errors, um, just because the natural thing to do is to treat it like any other agent. But I think that we’re gonna adjust and adapt very quickly. Um, I’ve been talking about this a bit in relation to the, uh, to the images. So, you know, like mid journey, uh, uh, photographic imagery now. I mean, it seems like this has happened in the last five minutes. Um, <laugh>, you know, in terms of how fast this is going.
Nick 01:02:28 So you’d better get this podcast up, uh, quickly before everything goes outta date, <laugh>. But, but now we’re getting these what seemed to be incredibly, you know, uh, realistic looking photographs, um, that can be generated by ai. And there’s, there’s all this panic about, um, oh, we’re going to be constantly tricked by deep fakes and that kind of thing. Yeah. Um, but you know, I, I was saying to someone the other day, I, what has to happen is a simple psychological shift that we no longer look at a photograph and go, oh, that’s a real thing that happened with someone, pho photo, photo photographing it. But rather, let’s say I, I did a little cartoon of, uh, Donald Trump getting arrested mm-hmm. <affirmative>, and I showed it to you, and I say, look, Donald Trump got arrested. You’d be like, uh, so you say it, it’s your little cartoon.
Nick 01:03:19 Um, but if I, you know, I do a fake news website and I go, look, it is this, you know, wonderful looking photograph. You’re gonna get tricked much more easily. Um, but I think that’s gonna change very quickly, and it should. Um, and, you know, that’s where we will simply recalibrate and change from instantaneously taking, treating that thing as, as treating that image in this case, the way that we’ve treated photographs, you know, 50 years ago, um, to just, uh, changing how we process them. So that is something that happens at our end, right? Yeah. And that’s the kind of change, that’s the kind of adjustment that we’ll need to make with large language models, is really adjust at our end how we digest this kind of input.
Paul 01:04:06 But are, aren’t we? So what you’re saying and is, is in essence, we’re gonna become less gullible, and it seems like we’ve become more gullible over time in, in the modern day, you know, going back to the idea like of two people interacting, right? And they’re using language, uh, to sort of coordinate around some problem, you know, the dog in the backyard or something. Uh, and so I kind of picture a problem. Yes, there’s a, um, it’s not a symmetric flow of, uh, influence. There’s always like one person influencing the other more, right? I mean, in general. Um, but with an ai, I don’t, I worry, you know, where is that flow of influence? Right? Um, and, and how does that fit within coordinating around solving a social, um, problem? So that was kind of a three different topics to throw at you there.
Nick 01:04:57 Well, I think, I mean, uh, accountability is something that I would kind of often come back to. So if, if, if we are in the house and I say, Hey, there’s a dog in the backyard, uh, and then you go out and look and there isn’t one, or, or, or, you know, it’s a, a sheep mm-hmm. <affirmative> or something mm-hmm. <affirmative>, um, I’m accountable for what I’ve said to you. It’s like, there’s not a dog, it’s a sheep, or, you know, there’s no dog. What are you talking about? Yeah. Uh, so it seems like a trivial case, but it’s really important because I, in a regular interaction, I am accountable in that sense for, for what I, for what I’ve said that I make a claim, uh, you understand that you act on it. And if it turns out to be wrong, you know, you can say, well, you misled me, or That’s not accurate, or you lied.
Nick 01:05:41 Or, you know, um, why’d you say that there’s some sort of problem of inappropriateness? Um, and then because of it, you might update. Um, or there may be some, you know, that we update what our relationship or, or, or whatever you can imagine sort of all the consequences of the accountability of the things we say. And that’s why we tend not to kind of lie. That’s why we tend to be pretty conservative in how we speak to the people around us, um, because we’re trying to maintain these, these relationships. Um, something like a a, an AI or, or, or also maybe a news outlet’s completely impersonal in that sense. Um, so maybe with a news outlet, you do have a, an accountability in the sense that I can, uh, withdraw my subscription. I mean, there’s a kind of some cost that can be paid, um, but with something like a large language model, it’s this completely impersonal thing.
Nick 01:06:34 And, and, um, uh, you, you just don’t have that kind of coordination that is like, let, let’s put it another way. It doesn’t really have any skin in the game. I mean, you are just kind of using it in some, in some way. Um, so maybe that addressed part of what you’re saying. I think the other thing is, is when we talk about gullibility y you know, that’s, it’s a very sort of, um, that I think kind of draws on this conduit metaphor that, you know, there’s a, there’s a message here. It comes over to me. Yeah. I unpack it. Do I, do I, do I accept it or not? Kind of thing. Um, and I think there’s much more, I keep trying to emphasize this with the relation between the sort of sender and receiver is not, it sending and receiving is not the best way to think about it because it’s really about a joint activity.
Nick 01:07:30 Um, so I tend to think ab, you know, I’m using more and more of the term calibration that you’re kind of getting calibrated to each other. Mm-hmm. <affirmative> and language, of course is a big technology for, for calibration and, and, uh, you know, that’s how it works. And we ha that’s what, you know, we have to learn it and then we can save all this time by kind of, cuz we’re all in inter calibrated in within the community to some tolerable degree. Um, so learning a new medium, uh, is, is like knowing a new person or going to a new culture in some sense that you are, you gotta get calibrated and, um, we just, you know, we’re not calibrated very well yet. I I wouldn’t think to the functionality of let’s say a lot of ais. Um, but we will be soon because, you know, a lot of us will make errors and, and, and update our understanding and, and, um, figure out how to, how to benefit
Paul 01:08:25 It is interesting. You, you said meeting new people and, and calibrating with the new person. So I meet Jessica and she seems real nice and I want to be her friend, so I’m more prone to calibrate, uh, with Jessica through language, et cetera, but I won’t have the choice with, uh, an ai. Right. And, and they’re, because they’re gonna be a very certain way. I mean, I don’t know how the varieties of quote unquote personalities that will be, um, you know, do you get to choose what kind of personality you want to interact with with your chat, uh, G P T or chat, um, bot ai, and then how will that affect, so let, let’s say, and, and we can bring the in things like Siri, right? The, the, so a lot of what you taught right about in the book is not just written language, but spoken language and the dynamics of our speech.
Paul 01:09:13 And so maybe we will come back to that, but, you know, things of that nature when people talk to Siri, Siri doesn’t adjust the way she speaks to you based on your emotions or how excited you are about a concert that you want to get tickets to or something. And I would, I would assume that, um, an AI like a chat, um, written language based, um, application also would, you know, does, doesn’t co calibrate, you’re not co calibrating because, so then you’re always calibrating to it. It has no ca accountability, no skin in the game. Right? A and I also just wonder how that’s gonna affect, if, if we are be gonna become more skeptical of a fake, uh, picture or fake text or something, will that affect how our skeptic, our our skepticism with other live agents as well? You know, how is that gonna affect the dynamics of society and language and so on?
Nick 01:10:08 <laugh>, it, it seems unlikely that it would affect, uh, you know, our calibrate you know, our, our understanding of live agents. I mean, people are really smart and, and very discerning and very much able to move. You know, we’re a kind of fish and fusion species in, in a radical way. We, we, we can create little groups, then we break and then form other groups, then we break and form other groups. We do it all throughout the day constantly. So every time we like move from, you know, you have a meeting over in this room with these people and then you go to have lunch with someone, um, you are changing your calibration. And it is very, very subtle, uh, kind of ways. Um, and so we’re adept at that. We’re good at that. And I think we’re very good at kind of learning how to shift, uh, between diets and, uh, and updating, you know, what we want things to mean and what we regard as being appropriate and inappropriate in this kind of context.
Nick 01:11:06 Um, in terms of, you know, let’s say flexibility of calibration, uh, I think what you say is true about certain, I don’t know, sir, something like that. Uh, it’s just always the same and it kind of has no emotion and all the rest of it. Um, there’s certainly, I think a lot of work out there going on that aims to achieve this kind of calibration with ai. So, I mean, there’s an app, um, I think it’s called Replica, um, where you set up a, um, uh, you know, a friend and, um, then it, you know, there’s, it’s cumulative. So this friend gets to know, you know, you chat with it all the time and it, you know, it, it, you build up a relationship, so it’s kind of like a, a fake friend or whatever. Yeah. Um, and you can very much specify who it is, what they’re like, what they look like, um, and then they learn who you are. And, um, and over time you build up this relationship. So it’s, a lot of people find that creepy and I can understand why, um, because deep down, you know, it isn’t another person in some sense. I mean it, I dunno, it’s maybe like a pet or something like that. So it’s good for some people and other people don’t really like it. Um, and probably even me saying that, uh, pet lovers are kind of like <laugh>. I know.
Paul 01:12:23 Yeah.
Nick 01:12:24 Be turned off, you gotta
Paul 01:12:25 Get in trouble. Yeah.
Nick 01:12:27 <laugh> saw your puppy dog in the background. Yeah. Um, just before. Uh, but I think what something that’s important is, you know, transparency and, and sort of an appropriateness to what we actually know about the nature of this, uh, of this agent. So thinking for example, about, uh, navigation, um, bots in cars, right? So you’re in your car, you’ve got your phone there, and it’s telling you, uh, want to turn left and turn right, things like that.
Paul 01:12:59 And you’re talking
Nick 01:13:00 To it. So sometimes What’s that?
Paul 01:13:02 And you’re probably talking to it, saying, I don’t wanna go that, you know, using language to talk. Yeah.
Nick 01:13:06 You could be talking to it. It’s not, it doesn’t care what you say. Yeah. Uh, but what I was thinking about was the relation between its instructions and your actions. So, um, the, let’s say, so sometimes if I’m going to, let’s say, go to point A, uh, in the car, uh, I dunno where point A is exactly, I dunno how to get there Exactly. So I put it into my phone, uh, but first I have to drop off someone at, you know, at some other place that I do know the way to. So the first part of the journey, I keep taking turns that are not the ones that the navigation is telling me to take. Mm-hmm. <affirmative>, it’s just gonna go left. But instead I go, right, and then it has to redo its little instruction and it says, okay, go left now. But I go straight ahead. Um, and all it does is it just silently recalculates and gives me a new instruction. But of course, if it was a real person sitting in the seat next to me, it’d be going, Hey, I said, turn left, what are you doing? Um, I said, go straight ahead. What, what’s up? Um, so it would be questioning me. It would be, and then it would be getting mad at me probably and say, you’re wasting my time. You know? Um, or are you okay? It may be actually now apps, <laugh>, they,
Paul 01:14:14 Yeah.
Nick 01:14:15 Yeah. I fell over the other day. My watch said, uh, should I call, uh, the emergency? Oh, really? Line. Oh, yeah. So, you know, you get, you get these, um, uh, maybe now at some point apps will say, you know, have you been on drugs? Or something like that? Hey, you’re having
Paul 01:14:28 A stroke. Yeah. But,
Nick 01:14:30 But my point is that if you, that I am, I don’t really want the app to be saying those things to me. Um, I know that it’s just a robot and I know what I’ve programmed it to do, and I don’t care that it has to recalculate the, the route on the way because I’ve taken a different turn. And so the transparency there is, is clear, you know, and I think there’s a lot of cases in which, because we tr we transparently know this is not a human being that we are accountable to, um, you know, that I don’t want to be held accountable to it. I don’t wanna be told, oh, you, you’re not taking the turn I told you to take. And if you started programming ais to do that kind of thing, people would be like, yeah, this either it creeps me out or it annoys me, or, you know, so I think this is all part of this, the, the, the idea that humans are excellent at relationships.
Nick 01:15:21 Um, we’re excellent at calibrating our understanding and also calibrating things like what’s required of us and what’s appropriate. Um, so, you know, take just in everyday life, forget computers and stuff, you’d go back 5,000 years, you’d have the same thing. Um, with certain people in our lives, we can talk about certain topics and with other people, we wouldn’t talk about those topics, right? So if something like, uh, I don’t know, um, uh, medical condition or something like that, um, you know, if you’re talking to a total stranger, you’re probably not very likely to start talking about that. Um, then if you are talking to your, uh, immediate family member, so we’re, we’re, we’re very good at, um, at, at, at, at calibration. And I don’t think that UMIs should, um, you know, pose any threat to that. We just need to, to learn, um, the different agents that are now in our landscape
Paul 01:16:19 For the very end of your book, um, you make the claim that we need to essentially strive to get our language right, um, because all language is bias and, um, influence and framing, um, a should we, so that has a normative flavor, and is that the way language works and or should work? Should we be trying to engineer the right kind of language? And the reason I’m asking this now is because I’m also just wondering about the effect that, um, AI generated language will have on our society and culture and language moving forward.
Nick 01:17:00 Yeah. When you say, should we be trying to engineer, um, you know, there’s, there’s different scales or different levels at which you could ask that question. So, uh, if you mean, should we be trying to engineer language, regulate it, uh, you know, get someone in the government to make decisions about how to say things, I think the answer is no, because like, who is this person gonna be? And, and, and whose interests are they going to be, um, representing? And, you know, will they be making good, good decisions? Um, so for me, if the question of engineering is about something like regulating, um, I, I think no, uh, because going back to the, the gatekeeper point I made earlier on, um, it’s really about us as individuals being mindful. I mean, you can’t, mindfulness is for minds. It’s not something that you can kind of make happen at a level far above that, or any level above that.
Nick 01:18:04 So the engineering that has to happen, and yes, engineering must happen, it’s at the level of us as individuals. So as a producer of language engineering would be, uh, just deciding what to say and how to say it. Um, that’s what we do every time we open our mouths. And part of the kind of mindfulness that I want to advocate is about just, just being aware of that and thinking about that and thinking, well, okay, I’m gonna say something. What’s the right way to say it? Now, sometimes we do, I mean, I think all of us have the experience of consciously thinking about that. How am I gonna say this thing? I wanna say, what are the best words to use? But oftentimes, um, you know, just the words just come out. Uh, so as a producer, I think we can be more mindful than we are, and it’s always about picking the right balance.
Nick 01:18:58 You simply can’t kind of overthink everything. And I, you know, wouldn’t be, uh, efficient wouldn’t be useful. The other part though, of engineering, uh, and I don’t know if engineering really applies in this, uh, this other part, but the other piece of this that’s just as important is from the comprehension side. So when a bit of language comes in, um, you know, you don’t take it to face value. Just like you, you no longer want to take a video or a photo on a website at face value. You want to think, okay, um, they said it like that. Why did they say it like that? Um, now when I put it like that, it seems a cynical, it seems kind of like, uh, untrusting in some way, and it, that that’s not how it, you know, sometimes it might be like that, but that’s not, this can be a very cooperative way to think as well.
Nick 01:19:46 So, you know, it’s about thinking, what does this person want? Is that something that I also want? Um, so oftentimes miscommunication is about missing those cues. It’s about not seeing, uh, uh, why someone said something in that way. And the, the clues come from things like the, the, the available choices that they had. So we always have different ways of saying the same thing. Um, and we refer to that difference as this difference in framing. And I think that being more aware of framing, being more sort of literate in some sense, and mindful is, uh, is, is just a crucial rule of thumb. And it’s really, to me, one of the most important aspects of, uh, of, of cognitive literacy. When people talk about things like media literacy, I think that’s ultimately what they, what they are talking about. And it’s goes far beyond media. It’s about really just, uh, uh, communication of, of, of, of any kind. Hmm.
Paul 01:20:48 Going back to the, you know, you, you, you talk at length in the book about spoken language as well, like I, uh, alluded to before, and one of the things that you write about is, um, Dunbar’s concentric circles. So, which I didn’t know about. Um, so there’s Dunbar’s famous number, is it 150? I forget. Uh,
Nick 01:21:07 Yeah, that’s one 50,
Paul 01:21:08 Yeah. About the number of people you can actively maintain in your social group. But the concentric circle idea is that you have around seven that you regularly communicate, um, on a, you know, on a regular basis and ask how their health is and things like that. And then, you know, there’s a little bit, the next circle out has more people in it, and you have a kind of a different relationship with them. You know, you exchange, uh, a narrower range of words. And the example that you give in the book is, um, one of the examples is a party where, uh, the person who answered the door was being recorded. And when I think her name, I don’t know what her name was, let’s say it was Sally, when Sally answers the door, and it’s her friend Beatrice, she goes, hi. And, you know, it’s like really long.
Paul 01:21:55 And Beatrice kind of says hello over, you know, over the top of her. And then when, later when, uh, Jim comes to the door and she doesn’t know Jim very well, she says hello, like, kind of much quicker, you know, and those relate to where you are on, on within the hierarchy of the concentric circles within Dunbar’s, um, number. Um, and it made me again think about, you know, Siri and how Siri is, you know, always, I guess toward the outermost ring would be in the outermost rings of those numbers. And I suppose what you’re saying is that we shouldn’t worry too much because we won’t, our interaction with these kinds of gadgets, whether it’s spoken or written, um, doesn’t, won’t really interfere with that kind of, uh, cognition, that social cognition of ours.
Nick 01:22:44 No, not as long as we have a life. Um, you know, I think that, uh, uh, Dunbar’s point is that, you know, any human is in a social network. And within that social network, I mean, the network is, is made up of, of, um, relationships. So, you know, datic relationships fundamentally. Um, and so you should, any individual person should be able to count up who are the people that they have a relationship with, and then you can characterize differences between those kinds of relationships. And that’s part of what the Dunbar’s work for many years has sort of been mapping out. And that is that, um, you can, you can create these kind of groups, uh, that, uh, that you have in your life. And those people, there are different implications for how you would act, uh, in accordance with, in with those people. That’s kind of what you were summarizing just just before, language plays a big role in that.
Nick 01:23:47 If you don’t maintain those social groupings, um, you know, then you really can’t survive in some important sense. I mean, the, the, you know, you literally, the, the innermost circle of, of your social world is, is very often defined with things like, you know, those are the people who would, um, bail you out. Uh, if there was a, a, a if you had a real problem, um, those are the people that would drop everything and, and go to help you if you had a health crisis. Um, things like that. And so that’s not just a fun way of defining who are the people in your inner circle, those very, that’s a very real, um, uh, consequence of being in someone’s inner circle. So you have certain people who are that to you, and you are that to certain other people. So that, that really is about accountability.
Nick 01:24:37 So you, you being in that circle and having entitlements, uh, with respect to others in that circle also means that you have obligations. Uh, and that’s what relationships are, are all about. And when it comes to things like Siri, well, you know, what are your entitlements? They’re very, very different in kind. I mean, your entitlement is that you can posee a question to Siri anytime, and, uh, it doesn’t matter what it is, and they’ll just answer it. You don’t have to say hello. You don’t have to say goodbye. There’s all these different rules around how you interact. Um, but people in your real life that that’s not possible, can’t ring someone up at three in the morning and say, you know, how long would it take me to drive to New York? And then they tell, they, you know, it wouldn’t, wouldn’t, it wouldn’t work. So they’re just a whole e every one of those concentric circles of Dunbars is about a set of entitlements and obligations in social relationships. And if, you know, and, and language has everything to do with enacting and defining what those kinds of entitlements and obligations are. And, uh, artificial agents are just, you know, a very different kind of beast when it comes to our social relationships. Hmm.
Paul 01:25:46 All right. Well, I really appreciate I, despite my, uh, scatterbrained questions, the, the book, I highly recommend it, and it’s much more organized than my questions were today. So, uh, <laugh>, thanks. I appreciate, very interesting. Okay, great chatting. Well, I appreciate it, Nick. Thanks for being on.
Nick 01:26:01 No problem. Thanks for having me.
Speaker 2 01:26:18 I alone produce brain inspired. If you value this podcast, consider supporting it through Patreon to access full versions of all the episodes and to join our Discord community. Or if you wanna learn more about the intersection of neuroscience and ai, consider signing up for my online course, neuro ai, the quest to explain intelligence. Go to brandin inspired.co. To learn more, to get in touch with me, email Paul Brennan inspired.co. You’re hearing music by the new year. Find email@example.com. Thank you. Thank you for your support. See you next time.