WEBVTT - How does language shape the way we think? 0:00:00.009 --> 0:00:02.990 This podcast was made on the lands of the Wurundjeri people, 0:00:03.240 --> 0:00:06.159 the Woi Wurrung and the Bunurong. We'd like to pay 0:00:06.159 --> 0:00:09.238 respects to their elders, past and present, and emerging. 0:00:14.560 --> 0:00:17.879 From the Melbourne School of Psychological Sciences at the University 0:00:17.879 --> 0:00:20.360 of Melbourne, this is PsychTalks. 0:00:22.329 --> 0:00:25.729 Hello and welcome to another round of PsychTalks. I'm Nick Haslam, 0:00:25.819 --> 0:00:28.250 And I'm Cassie Hayward. We're your hosts, and we're just 0:00:28.250 --> 0:00:31.329 itching to deep dive into more fascinating research in psychology 0:00:31.329 --> 0:00:35.380 and neuroscience. This season of PsychTalks is already halfway, so 0:00:35.380 --> 0:00:37.610 if you haven't already, make sure you subscribe so you 0:00:37.610 --> 0:00:38.930 don't miss any episodes. 0:00:39.459 --> 0:00:42.259 Today we're talking with Dr Frank Mollica. Frank studies the 0:00:42.259 --> 0:00:44.659 wonders of human language, what it tells us about how 0:00:44.659 --> 0:00:47.330 we think and how we interact with the world around us. 0:00:48.110 --> 0:00:51.269 His research crisscrosses numerous cultural and linguistic settings, and it 0:00:51.270 --> 0:00:53.229 also has a lot to say about the horror of 0:00:53.229 --> 0:00:55.189 legal language. Let's get into it. 0:01:01.549 --> 0:01:02.330 Welcome, Frank. 0:01:02.590 --> 0:01:04.039 Hi, thanks for having me. 0:01:04.190 --> 0:01:07.620 So, Frank, you describe yourself as a computational cognitive scientist, 0:01:07.669 --> 0:01:09.989 and now that might be unfamiliar to some of our listeners. 0:01:10.029 --> 0:01:12.470 So what does it mean? And can you explain what 0:01:12.470 --> 0:01:15.940 cognitive science is and what's computational about your approach to it? 0:01:16.110 --> 0:01:21.510 Yeah. So, cognitive science is an interdisciplinary field. It mixes anthropology, 0:01:21.750 --> 0:01:26.419 computer science, linguistics, neuroscience, philosophy, and of course, psychology. Uh, 0:01:26.480 --> 0:01:27.900 we study the mind like a computer. 0:01:28.779 --> 0:01:31.779 Right, uh, what are the mental representations and processes, the 0:01:31.779 --> 0:01:34.849 computer programme basically, that explains human reasoning? 0:01:35.500 --> 0:01:38.699 So I'm a computational cognitive scientist, I spend most of 0:01:38.699 --> 0:01:42.019 my time translating between these two fields and basically taking 0:01:42.019 --> 0:01:43.779 the theories and the insights from these fields and putting 0:01:43.779 --> 0:01:46.529 them into math, and then, you know, formalising the math, 0:01:46.540 --> 0:01:49.699 making nice testable predictions that, you know, I then go 0:01:49.699 --> 0:01:51.959 out and find collaborators or go into the field and 0:01:51.970 --> 0:01:53.970 and test with large data normally. 0:01:54.319 --> 0:01:57.449 Um, as you've said, the focus of your research is language, 0:01:57.819 --> 0:02:01.019 but you talk about it as a cognitive technology. What 0:02:01.019 --> 0:02:02.660 do you mean by that? OK. 0:02:03.209 --> 0:02:06.500 So, a cognitive technology, I guess at the simplest level, 0:02:06.589 --> 0:02:09.410 it's a tool. For example, uh, language and number are 0:02:09.410 --> 0:02:12.850 cognitive technologies, right, we invented that, that's our fault, right, 0:02:12.970 --> 0:02:15.369 like we have littered the world with linguistic structure and 0:02:15.369 --> 0:02:17.889 numerical structure, and then we learn from it and we 0:02:17.889 --> 0:02:19.169 use it to achieve these goals. 0:02:19.529 --> 0:02:23.070 Um, and importantly with these cognitive technologies, we learn them, right? 0:02:23.149 --> 0:02:25.990 We learn them as kids, we use them to, you know, 0:02:26.229 --> 0:02:28.589 achieve different goals, whether it's, you know, math or whether 0:02:28.589 --> 0:02:32.229 it's communication, right, and that leaves extra structure in the 0:02:32.229 --> 0:02:36.110 world that other people learn from. And additionally, we teach 0:02:36.110 --> 0:02:39.769 these things, right, we explicitly teach people number, we explicitly 0:02:39.770 --> 0:02:41.509 teach people, right, like language. 0:02:41.955 --> 0:02:45.585 Right, and through this repeated pattern of uh learning and teaching, 0:02:45.794 --> 0:02:49.175 uh we culturally evolve optimal solutions to our problems, right, 0:02:49.235 --> 0:02:51.954 that help us to achieve our goals, and this actually 0:02:51.955 --> 0:02:54.354 allows us to flexibly adapt to the environment in ways 0:02:54.354 --> 0:02:56.445 that biological evolution wouldn't allow us to. 0:02:56.595 --> 0:02:58.675 This is so interesting because the idea that language is 0:02:58.675 --> 0:03:02.154 shaped by culture is sort of intuitive to most of us, um, 0:03:02.195 --> 0:03:03.554 the idea that, you know, different, 0:03:04.119 --> 0:03:06.418 Languages split up the world in different kind of ways. 0:03:06.440 --> 0:03:08.500 But I remember back in the dark ages when I 0:03:08.500 --> 0:03:11.979 was studying language, uh, the emphasis was, was much more 0:03:11.979 --> 0:03:15.500 on universals, on how all languages are in, if you like, 0:03:15.660 --> 0:03:17.130 built from the same building blocks. 0:03:17.460 --> 0:03:21.259 So can you give us an example, um, say, colour words, 0:03:21.380 --> 0:03:24.149 how do different cultures or languages, if you like, um, 0:03:24.339 --> 0:03:25.570 break up the colour space? 0:03:25.800 --> 0:03:29.258 Yeah, so colour is this fun case, right, we all 0:03:29.258 --> 0:03:32.220 can perceive the same colours if we have normal vision, right, 0:03:32.258 --> 0:03:35.490 we can distinguish, you know, the eggshell white from the white, um, 0:03:35.500 --> 0:03:37.850 but different languages of the world, they don't have a 0:03:37.850 --> 0:03:40.809 vocabulary that's that rich, right? Um, the basic 0:03:41.110 --> 0:03:43.490 colour terms, the ones that we use every day, right, 0:03:43.610 --> 0:03:48.289 not the eggshell, off-white, mauve, turquoise, right? There's only a 0:03:48.289 --> 0:03:52.210 finite inventory of these basic colour terms. Uh, English has 0:03:52.210 --> 0:03:56.690 about 11, right, um, blue, yellow, pink, purple, um, but 0:03:56.690 --> 0:03:59.660 other languages have fewer and some languages have more. So 0:03:59.660 --> 0:04:02.529 for example, Agarabi, uh, a language they speak in Papua 0:04:02.529 --> 0:04:05.360 New Guinea, uh, that only has about 5 colour terms that, 0:04:05.410 --> 0:04:08.490 you know, everyone would use, right, they use like blue, green, yellow, red, 0:04:08.490 --> 0:04:09.610 and dark or black. 0:04:09.960 --> 0:04:14.279 Um, whereas other languages like uh Mexican Spanish, for example, 0:04:14.529 --> 0:04:18.130 has two blues, uh, they have a celeste kind of 0:04:18.130 --> 0:04:20.808 sky blue, and then the rest of what an English 0:04:20.809 --> 0:04:21.730 person will call blue. 0:04:22.119 --> 0:04:25.359 Uh, similarly, Russian also has two blues, except they carve 0:04:25.359 --> 0:04:27.720 it up differently. They have like the blue that you know, 0:04:27.920 --> 0:04:30.320 generally English people would, you know, agree is blue, but 0:04:30.320 --> 0:04:32.440 then they carve off a bit that's like a navy blue, 0:04:32.480 --> 0:04:33.269 it's darker. 0:04:33.720 --> 0:04:36.440 So interesting, so I guess part of what a computational 0:04:36.440 --> 0:04:39.559 cognitive scientist might do is see whether there are principles 0:04:39.559 --> 0:04:42.600 underlying these differences, uh, right? So, uh, one of the 0:04:42.600 --> 0:04:44.760 ones you refer to, I think when you talk about 0:04:44.760 --> 0:04:45.359 this stuff is, 0:04:45.700 --> 0:04:48.459 a principle of efficiency. So what do you mean by 0:04:48.459 --> 0:04:51.829 efficiency and how do you show that languages are efficient 0:04:51.829 --> 0:04:52.380 or not? 0:04:52.630 --> 0:04:55.669 Yeah. So this is the day job, right? Figuring out 0:04:55.670 --> 0:04:58.230 what are these underlying principles that can explain like all 0:04:58.230 --> 0:05:01.469 of this variation. We've optimally solved these goals, right? We, 0:05:01.630 --> 0:05:04.549 we've come up with these efficient structures like language or number, right? 0:05:04.630 --> 0:05:06.790 But how do we get there? Uh, and so one 0:05:06.790 --> 0:05:09.190 thing I want to clarify is that goals don't always 0:05:09.190 --> 0:05:10.390 go in the same direction. 0:05:10.970 --> 0:05:13.730 Right, so let's take for example, a listener, a listener's 0:05:13.730 --> 0:05:15.809 goal might be to hear a word and then be 0:05:15.809 --> 0:05:19.010 able to identify any object in the world, right? So 0:05:19.010 --> 0:05:20.850 if I give you a word that can point to 0:05:20.850 --> 0:05:23.929 exactly the object that's intended, an optimal language from a 0:05:23.928 --> 0:05:27.410 listener's perspective is a language where every possible state of 0:05:27.410 --> 0:05:30.890 the world has its own unique word, right? So that 0:05:30.890 --> 0:05:34.529 means we have a unique word for all 621 Marvel 0:05:34.529 --> 0:05:36.969 universes and every single item inside them. 0:05:37.500 --> 0:05:40.859 Right, uh, impossible, we can't do that, especially because if 0:05:40.859 --> 0:05:43.299 you think about what a speaker's goal is, a speaker 0:05:43.299 --> 0:05:45.019 wants to be able to choose the word as quickly 0:05:45.019 --> 0:05:47.118 and accurately as possible, that's going to get their listener 0:05:47.119 --> 0:05:50.450 to understand, right? They don't want to search memory through 0:05:50.450 --> 0:05:52.660 an infinite, you know, amount of words in order to 0:05:52.660 --> 0:05:54.558 figure out what's the right word that would get my 0:05:54.559 --> 0:05:56.700 speaker to understand what I'm talking about. They have only 0:05:56.700 --> 0:05:59.299 a finite memory, they want us to be quick. A 0:05:59.299 --> 0:06:03.488 speaker's optimal language would essentially be one word, right, 'ba', 0:06:03.678 --> 0:06:05.820 and when I say 'ba', it means whatever I intended 0:06:05.820 --> 0:06:06.618 a speaker to mean. 0:06:06.950 --> 0:06:09.179 So all of my intentions, just one word, ba, I 0:06:09.178 --> 0:06:11.579 don't have to think about it, right? Now, of course, 0:06:11.660 --> 0:06:14.220 neither of these languages actually work, right, we call them 0:06:14.220 --> 0:06:17.380 degenerate languages, but real languages have to figure out how 0:06:17.380 --> 0:06:20.190 to trade off these two goals and how they choose 0:06:20.190 --> 0:06:24.768 to budget these goals defines a sort of efficiency trade-off. 0:06:25.309 --> 0:06:28.229 Right, it defines a trade-off between these two pressures, one 0:06:28.230 --> 0:06:31.070 for uh simplicity, we don't want a language that blows 0:06:31.070 --> 0:06:34.269 up in like a vocabulary, right, but also informativity, we 0:06:34.269 --> 0:06:36.070 want all of the new words that we add to 0:06:36.070 --> 0:06:38.470 be useful, right, to help us identify the things that 0:06:38.470 --> 0:06:41.339 we care about. Uh and so languages of the world 0:06:41.339 --> 0:06:45.308 tend to efficiently optimise this trade-off, right? Each language gets 0:06:45.309 --> 0:06:47.779 to choose how it wants to budget and spend on, 0:06:47.829 --> 0:06:50.950 you know, uh this much complexity for this much informativity. 0:06:51.390 --> 0:06:54.140 And what's interesting is we can build these computational models 0:06:54.140 --> 0:06:56.619 off of these two principles that define all of the 0:06:56.619 --> 0:07:00.940 different ways that you could budget between informativity and uh 0:07:00.940 --> 0:07:04.140 and simplicity. And when you do that, you get a 0:07:04.140 --> 0:07:07.940 whole variety of sort of a trade-off Pareto frontier of 0:07:07.940 --> 0:07:11.100 possible ways that languages could be, right? They could solve 0:07:11.100 --> 0:07:11.660 this problem. 0:07:12.010 --> 0:07:14.769 And all of the existing languages for something like colour 0:07:14.769 --> 0:07:19.500 tend to fall exactly near this boundary, suggesting that every 0:07:19.500 --> 0:07:24.019 language is indeed efficiently solving this problem. They're balancing their goals, 0:07:24.220 --> 0:07:27.059 but they all pick how they want to budget, you know, uh, which, 0:07:27.140 --> 0:07:30.820 which goal do I prefer more or prioritise more differently. 0:07:31.070 --> 0:07:33.519 And that actually characterises the diversity that we see in 0:07:33.519 --> 0:07:33.950 the world. 0:07:34.160 --> 0:07:38.239 Is it just about efficiency, or does it shape how 0:07:38.239 --> 0:07:39.920 we see the world? I had to chuckle when you 0:07:39.920 --> 0:07:42.640 talked about the Marvel Universe because I feel like I've 0:07:42.640 --> 0:07:44.440 learned a whole new language over the past few years 0:07:44.440 --> 0:07:46.679 as my kids become obsessed with soccer. So I've learned 0:07:46.679 --> 0:07:49.720 a whole new words, new everything. 0:07:50.660 --> 0:07:53.529 Am I changing the way I see the world with that? 0:07:53.940 --> 0:07:56.899 So I would argue that language definitely changes how you 0:07:56.899 --> 0:07:59.700 see the world, but it's not in like the Arrival, 0:07:59.739 --> 0:08:01.380 the movie kind of way where if you learn the 0:08:01.380 --> 0:08:04.380 alien language, right, suddenly you can time travel, right? And 0:08:04.380 --> 0:08:07.339 it's also not in the perceptual kind of way, where like, 0:08:07.440 --> 0:08:10.019 you know, my language has two colours for two colour 0:08:10.019 --> 0:08:12.730 words for blue, now I can, you know, see blues differently, 0:08:12.869 --> 0:08:14.500 because we can all see, you know, the shades of 0:08:14.500 --> 0:08:16.739 blue is the same, right, we can all distinguish between them. 0:08:17.179 --> 0:08:20.470 Um, but what language does do is it points attention 0:08:20.750 --> 0:08:22.470 on all of the structure in the world and it 0:08:22.470 --> 0:08:25.989 highlights what's important, right, what's going to be useful distinctions 0:08:25.989 --> 0:08:29.510 for later things in learning, right? Language is a primary 0:08:29.510 --> 0:08:33.189 tool for social transmission, right, this core thing, uh, where 0:08:33.190 --> 0:08:36.859 if we're going to culturally evolve good optimal solutions for, 0:08:36.869 --> 0:08:38.549 you know, our different goals, then we need to be 0:08:38.549 --> 0:08:42.109 able to transmit it. Uh, so for one example, to 0:08:42.109 --> 0:08:43.710 give you, you have to make this concrete, is if 0:08:43.710 --> 0:08:45.380 you think about like kinship terms. 0:08:46.059 --> 0:08:50.030 Right, kinship terms, you know, everybody has a family tree, right, 0:08:50.460 --> 0:08:52.959 uh kinship terms are things like mother and brother and whatnot. 0:08:53.950 --> 0:08:56.468 So everyone has a family tree, but it's invisible, you 0:08:56.469 --> 0:08:58.950 don't see it, right, like it's latent. If you're a 0:08:58.950 --> 0:09:00.630 kid and you're trying to figure out what your kinship 0:09:00.630 --> 0:09:03.659 terms are, you have to figure out basically just from 0:09:03.659 --> 0:09:07.900 the words as highlighting different people, what the underlying relationships are, 0:09:08.109 --> 0:09:12.150 and these underlying relationships can have different goals that they're 0:09:12.150 --> 0:09:14.150 made to achieve that you don't actually pay attention to 0:09:14.150 --> 0:09:17.699 or even see when you're a kid. For example, uh, 0:09:17.710 --> 0:09:20.468 we can think about the Yanomami tribe, uh, these are 0:09:20.469 --> 0:09:23.669 people that live in the Amazon rainforest, right? 0:09:24.280 --> 0:09:29.219 Whereas English doesn't separate uh cousins, your mother's brothers, uh, 0:09:29.320 --> 0:09:33.710 your father's brothers, your mother's sisters, your father's brother's kids, 0:09:33.919 --> 0:09:36.510 they all get the same term, they're all your cousin, right? 0:09:36.619 --> 0:09:40.239 And Yanomami, they actually care about uh whether you're a 0:09:40.239 --> 0:09:43.440 parallel cousin or a cross cousin. So a parallel cousin 0:09:43.440 --> 0:09:45.000 is your mother's sister's kid. 0:09:45.320 --> 0:09:48.439 Your father's brother's kids, right? A cross cousin is your 0:09:48.440 --> 0:09:52.510 mother's brother's kids or your father's sister's kids, right? Um, 0:09:52.520 --> 0:09:54.429 and so kids at a very young age have to learn, 0:09:54.599 --> 0:09:58.429 you know, the difference between these two, different groupings, right? Um, 0:09:58.440 --> 0:10:00.400 and this for a kid, you know, it can seem 0:10:00.400 --> 0:10:03.520 very arbitrary, it's a nice genealogical relationship, it exists, you 0:10:03.520 --> 0:10:05.919 can look at it on the tree, right? But like, why? 0:10:06.299 --> 0:10:08.630 Um, and it turns out that these kind of relationships 0:10:08.630 --> 0:10:11.390 actually have more to do, uh, with in this case 0:10:11.390 --> 0:10:15.309 in mating, for example, where, uh, it's tough to find, uh, 0:10:15.349 --> 0:10:17.989 you know, a potential mate in these kind of tribal places, right, 0:10:18.229 --> 0:10:21.380 and so this community solves that by having preferential marriages 0:10:21.380 --> 0:10:24.349 to cross cousins when they're mate limited. So your cross 0:10:24.349 --> 0:10:28.159 cousins are potentially future mates, um, but this doesn't matter 0:10:28.159 --> 0:10:30.349 to a kid, right? Like a kid who's learning this 0:10:30.349 --> 0:10:32.468 language is not going, who am I mating, they're a kid. 0:10:33.159 --> 0:10:35.719 Right? Um, so without language, you wouldn't be able to 0:10:35.719 --> 0:10:37.979 be aware of these kind of distinctions like early on, 0:10:38.049 --> 0:10:40.859 and these kind of distinctions can highlight or build uh 0:10:40.859 --> 0:10:44.179 structures that, you know, are useful for other goals later 0:10:44.179 --> 0:10:44.809 on in life. 0:10:45.099 --> 0:10:47.510 That's so interesting and so you've given us an explanation 0:10:47.510 --> 0:10:50.789 for why the Yanomami might have a different, way of 0:10:51.330 --> 0:10:55.890 dividing up kinship than we do here. What's the reason why, 0:10:56.049 --> 0:10:59.650 just going back to colour, why Russians would divide blue 0:10:59.650 --> 0:11:04.869 in a different way from Mexicans and why we English 0:11:04.869 --> 0:11:05.819 folk don't? 0:11:06.250 --> 0:11:08.569 Uh, so this is a, a nice question. I don't 0:11:08.570 --> 0:11:11.570 know if I'm gonna have a satisfying answer, um. 0:11:12.260 --> 0:11:15.209 The answer that I would uh suggest or I'd go 0:11:15.210 --> 0:11:17.449 with a hypothesis at the moment is that if we 0:11:17.450 --> 0:11:21.320 want to communicate about things informatively, right, that means that 0:11:21.320 --> 0:11:23.429 we need to know how often we need to talk 0:11:23.429 --> 0:11:26.369 about or this particular tool in this case is word, 0:11:26.599 --> 0:11:29.400 is going to be useful in achieving our goal. Um, 0:11:29.409 --> 0:11:31.090 and so if you look at the, the different parts 0:11:31.090 --> 0:11:32.650 of the colour space, what we want to do is 0:11:32.650 --> 0:11:36.130 we want to normally identify things by colour. 0:11:36.719 --> 0:11:39.229 Right, and so if there are things that are important 0:11:39.229 --> 0:11:42.030 to identify by colour, right, that's how we're going to 0:11:42.030 --> 0:11:43.510 carve up our colour space so that we can be 0:11:43.510 --> 0:11:47.510 more precise and have harder edges around those specific cases. 0:11:47.750 --> 0:11:50.940 So sort of adapted to the local cultural or ecological 0:11:51.510 --> 0:11:52.900 or something environmental context? 0:11:53.489 --> 0:11:55.869 Ideally it would be adapted, uh, exactly, it would be 0:11:55.869 --> 0:11:57.030 adapted to the local context. 0:11:57.299 --> 0:11:59.219 So for example, if you look at uh places that 0:11:59.219 --> 0:12:03.140 are tropical, have lots of, uh, you know, rainforests, green vibes, 0:12:03.260 --> 0:12:06.260 lots of bright, vibrant colours, uh, they tend to split 0:12:06.260 --> 0:12:08.819 the colour space so that they, they make really clear 0:12:08.820 --> 0:12:13.059 these bright, you know, uh, colourful, poisonous amphibians that you 0:12:13.059 --> 0:12:15.820 should not touch, uh, versus, you know, things that are 0:12:15.820 --> 0:12:17.530 OK to touch and interact with. 0:12:18.020 --> 0:12:20.619 So another aspect of language you've studied, and you brought 0:12:20.619 --> 0:12:23.059 this up in relation to the uh kinship terms a 0:12:23.059 --> 0:12:25.619 little bit, is how language is acquired. 0:12:25.880 --> 0:12:28.739 And you were involved in some fascinating work again in 0:12:28.739 --> 0:12:32.570 South America on the development of mathematical, uh, knowledge and 0:12:32.570 --> 0:12:36.460 number concepts in particular indigenous people, uh, in that region. 0:12:36.609 --> 0:12:38.449 Can you tell us a bit about that work and 0:12:38.450 --> 0:12:39.319 what it showed? 0:12:39.609 --> 0:12:42.729 Yeah. So I was lucky enough to be able to 0:12:42.729 --> 0:12:45.250 collaborate with some people who are working, uh, in Bolivia 0:12:45.250 --> 0:12:49.049 with the Tsimane people. Uh, the Tsimane are hunter gatherers, 0:12:49.090 --> 0:12:51.679 it's a hunter-gatherer society, uh, they live not far from 0:12:51.679 --> 0:12:52.919 La Paz along a river. 0:12:53.440 --> 0:12:56.760 So what we're interested in is primarily numerical developments and 0:12:56.760 --> 0:12:59.159 what's fun is that the Tsimane language has a base 0:12:59.159 --> 0:13:05.679 10 counting system, right, so 1-2-3-4-5-6-7-8-9-10, then it sort of resets, right? 0:13:05.919 --> 0:13:08.439 A lot of other industrialised societies also have, you know, 0:13:08.520 --> 0:13:11.719 a base 10 counting system, right, any English two-year-old can 0:13:11.719 --> 0:13:13.239 sing that song for you, they know how to count 0:13:13.239 --> 0:13:15.349 like 1-2-3-4-5-6-7-8-9-10. 0:13:16.039 --> 0:13:17.929 The interesting thing though is that while a 2 year 0:13:17.929 --> 0:13:21.169 old can count that, 2-year-olds don't actually know what those 0:13:21.169 --> 0:13:23.489 words mean. So if you put a pile of cookies 0:13:23.489 --> 0:13:24.968 in front of the same 2 year old who just 0:13:24.969 --> 0:13:27.289 counted 10 for you and ask them, Can I have 0:13:27.289 --> 0:13:31.010 3 cookies, right, you're just as likely to get the 0:13:31.010 --> 0:13:33.570 entire pile of cookies as you are to get however 0:13:33.570 --> 0:13:36.010 many fit in their hand, and certainly not 3. 0:13:36.500 --> 0:13:41.020 Um, right, kids learn number words in stages. Uh, first 0:13:41.020 --> 0:13:42.739 they learn what the meaning of one is, they can 0:13:42.739 --> 0:13:45.099 hand you one cookie exactly, then they learn what the 0:13:45.099 --> 0:13:46.940 meaning of two is, right, so they can hand you 0:13:46.940 --> 0:13:49.140 1 or 2 cookies, but you ask them for 3, 0:13:49.169 --> 0:13:51.780 you still get that handful, then they become 3-knowers and 0:13:51.780 --> 0:13:55.020 they figure out what 3 means. Sometimes you see kids 0:13:55.020 --> 0:13:57.900 that are 4-knowers, so they know what 4 means, but 0:13:57.900 --> 0:14:00.780 what's really cool is that we don't have 5-knowers. 0:14:01.219 --> 0:14:04.500 Right, By the time that a kid would be a 5-knower, right, 0:14:04.659 --> 0:14:06.739 they just figured it out, they figured out the algorithm, 0:14:07.020 --> 0:14:09.739 they can now count. So however many numbers that they 0:14:09.739 --> 0:14:12.580 can count to, they can now accurately, well, as accurately 0:14:12.580 --> 0:14:15.939 as they can count, they can hand you that many cookies, right? 0:14:16.219 --> 0:14:18.859 What's really cool is this happens in all industrialised societies 0:14:18.859 --> 0:14:20.900 that have a base 10 counting system. It takes a 0:14:20.900 --> 0:14:22.979 couple of years, right, we see this in Japan, we 0:14:22.979 --> 0:14:25.190 see this in Hebrew, Arabic, um. 0:14:25.630 --> 0:14:28.190 Every language that we have data for, right, kids go 0:14:28.190 --> 0:14:32.260 through the stage of development, you know, uh 1-knower, 2-knower, 3-knower, 4-knower. 0:14:33.219 --> 0:14:36.469 We want to know, hunter-gatherer society, Tsimane kids, do they 0:14:36.469 --> 0:14:38.390 show the same pattern, right? They also have a base 0:14:38.390 --> 0:14:41.309 10 system, but you know, very different from these industrialised 0:14:41.309 --> 0:14:44.780 societies that we have data for. So, uh, you know, uh, 0:14:44.909 --> 0:14:47.340 my collaborators basically went and looked at the Tsimane children 0:14:47.340 --> 0:14:49.030 and we found out that yes, they do go through 0:14:49.030 --> 0:14:53.070 this exact same pattern. Uh, they go 1-knower, 2-knower, 3-knower, 4-knower, 0:14:53.390 --> 0:14:56.469 and then eventually they figure out the algorithm. What's really 0:14:56.469 --> 0:14:59.270 cool is we recently did a meta-analysis, and it turns 0:14:59.270 --> 0:15:02.030 out that the cardinal principal knowers, the kids that figured 0:15:02.030 --> 0:15:02.950 out the algorithm. 0:15:03.599 --> 0:15:07.619 They all have some formal schooling, um, and formal schooling 0:15:07.619 --> 0:15:10.500 among the Tsimane is actually really recent, um, and it's 0:15:10.500 --> 0:15:14.169 not formal schooling like an industrialised society, is much more like, um, 0:15:14.299 --> 0:15:17.260 you know, when the government can provide aid, they have 0:15:17.260 --> 0:15:19.580 a teacher who comes and they teach the Tsimane kids 0:15:19.580 --> 0:15:22.059 in Spanish, uh, and so the instruction is in Spanish 0:15:22.059 --> 0:15:25.700 and it's very recent, um, but only the kids with 0:15:25.700 --> 0:15:28.820 formal schooling, uh, have, at least in our data set, uh, 0:15:28.940 --> 0:15:30.820 have acquired that full counting ability. 0:15:31.510 --> 0:15:34.750 But what's really cool is that there are people, there 0:15:34.750 --> 0:15:37.460 are Tsimane that are adults that, you know, did not 0:15:37.460 --> 0:15:40.909 have any of this formal schooling, uh, and they can count, right? 0:15:41.059 --> 0:15:42.510 And so we wanted to figure out, you know, well, 0:15:42.630 --> 0:15:44.109 how did they figure it out, right? They didn't have 0:15:44.109 --> 0:15:47.789 formal schooling, so is formal schooling actually required, right? Um, 0:15:47.799 --> 0:15:50.669 and so the anthropologist on our team, David O'Shaughnessy, went 0:15:50.669 --> 0:15:54.299 and actually uh went to go find adults that, you know, 0:15:54.590 --> 0:15:56.789 might not have had the experience of formal schooling, but 0:15:56.789 --> 0:15:59.710 they might have had mathematical uh experience elsewhere, they might 0:15:59.710 --> 0:16:00.469 have done trade. 0:16:00.880 --> 0:16:04.479 Right, so the Tsimane do trade with uh other people, right, 0:16:04.669 --> 0:16:06.909 and the trade language is sort of Spanish because that's a, 0:16:07.080 --> 0:16:10.320 you know, the the lingua franca of Bolivia, right? Um, 0:16:10.760 --> 0:16:13.940 and what they found, Dave, when, when he went there, 0:16:14.000 --> 0:16:15.530 he found out that basically 0:16:16.289 --> 0:16:17.080 these people 0:16:18.070 --> 0:16:20.940 maybe they, they can't exactly count, but they can do trade, 0:16:21.070 --> 0:16:23.309 and so he found this really cool uh group of 0:16:23.309 --> 0:16:28.719 people who could do mathematical computations that were necessary for trade. 0:16:29.070 --> 0:16:32.270 So for example, uh, in the Tremane society, basically trade 0:16:32.270 --> 0:16:35.309 works in fives, you have these jatata leaves uh that 0:16:35.309 --> 0:16:37.869 they trade or sometimes they'll trade like a bunch of 0:16:37.869 --> 0:16:40.789 uh bananas or plantains if they're in the right sort 0:16:40.789 --> 0:16:43.869 of amount, like for a cluster. They basically group them 0:16:43.869 --> 0:16:45.669 in 5 and they do multiples of 5 is like 0:16:45.669 --> 0:16:46.380 the the trade. 0:16:46.729 --> 0:16:49.090 So, uh, if you ask these people, you know, 5 0:16:49.090 --> 0:16:52.080 times 2, 5 times 3, 10 times 4, fine, 10 0:16:52.080 --> 0:16:56.090 plus 5, fine, 3 plus 4, no idea, wrong, right, 0:16:56.210 --> 0:16:57.650 2 plus 5, nope. 0:16:58.309 --> 0:17:01.669 Um, right, so they figured out enough math, they figured 0:17:01.669 --> 0:17:04.430 out math that works specifically for them for the goals 0:17:04.430 --> 0:17:06.910 that they're trying to achieve, right? And I think that's 0:17:06.910 --> 0:17:10.160 why I guess numerical cognition is such a great idea, 0:17:10.390 --> 0:17:14.589 such a great example of these cognitive technologies, right, these 0:17:14.589 --> 0:17:18.069 technologies that are customly adapted to our goals, right, if 0:17:18.069 --> 0:17:20.069 the only thing that I really need to do is 0:17:20.069 --> 0:17:21.979 trade in multiples of 5, I'm going to learn math 0:17:21.979 --> 0:17:24.430 in multiples of 5, right? Like why learn that 1 0:17:24.430 --> 0:17:25.339 + 1 stuff? 0:17:25.780 --> 0:17:28.390 Um, and I mean, this isn't the first time an 0:17:28.400 --> 0:17:32.218 anthropologist has found a result. Um, if you look back, uh, 0:17:32.270 --> 0:17:35.069 Jeff Sachs has worked with Brazilian candy sellers during periods 0:17:35.069 --> 0:17:38.670 of inflation, where, uh, if you're trying to sell candy 0:17:38.670 --> 0:17:41.349 in massive periods of inflation, these like 6 to 10 0:17:41.349 --> 0:17:43.709 year old boys would basically be trying to sell candy 0:17:43.709 --> 0:17:47.459 at the public transportation, uh, stops, right, but you wouldn't know, 0:17:47.670 --> 0:17:49.468 like how much am I actually going to need to 0:17:49.469 --> 0:17:51.660 cost these things because the, you know, the, the price 0:17:51.920 --> 0:17:53.659 changes rapidly over the course of a day. 0:17:54.030 --> 0:17:56.069 Right, so you have to sort of really quickly offhand 0:17:56.069 --> 0:17:59.948 make some heuristics that in this case, use a magnitude 0:17:59.949 --> 0:18:03.020 comparisons to figure out how you should price your candy 0:18:03.020 --> 0:18:04.949 so that you have enough to buy tomorrow's box of 0:18:04.949 --> 0:18:06.630 candy and that you still make a profit. 0:18:07.229 --> 0:18:10.349 Right, and when you compare these, you know, like non-schooled 0:18:10.349 --> 0:18:14.119 Brazilian candy sellers to, you know, people of equivalent socioeconomic 0:18:14.119 --> 0:18:16.790 status with schooling or even the the kids that have, 0:18:16.869 --> 0:18:20.459 you know, like a good socioeconomic status and, you know, schooling, 0:18:20.869 --> 0:18:25.750 they're better at these abstract magnitude comparisons than their peers, right, 0:18:25.790 --> 0:18:27.669 they can't do the formal counting thing that they would 0:18:27.670 --> 0:18:29.069 teach you in school, but it's like they can teach 0:18:29.069 --> 0:18:32.699 you what they needed, right, and they're better at it, um, so. 0:18:33.209 --> 0:18:36.000 These kind of things really are adapted, cognitive technologies are, 0:18:36.050 --> 0:18:38.920 are really adapted to what we need to do. 0:18:39.380 --> 0:18:42.459 I think they're such fascinating examples of how efficient language 0:18:42.459 --> 0:18:44.930 can be, but also so adapted to the goals that 0:18:44.930 --> 0:18:47.849 you need in that, in that scenario. But you've also 0:18:47.849 --> 0:18:52.079 done research on the type of language that's spectacularly inefficient. 0:18:52.329 --> 0:18:54.349 Can you tell us about legalese? 0:18:54.660 --> 0:18:56.089 Yes, um. 0:18:57.050 --> 0:19:01.020 Legalese is notorious, right? What do I really have to say? Um, 0:19:01.280 --> 0:19:04.489 legalese is difficult in so many different ways, um, rather 0:19:04.489 --> 0:19:07.280 than just saying a clean sentence, legalese likes to take 0:19:07.280 --> 0:19:08.810 a sentence and throw it in the middle of the 0:19:08.810 --> 0:19:13.250 other sentence, right? For example, the contractor who knowingly and 0:19:13.250 --> 0:19:16.968 in sound mind and good conscious entered this employment agreement 0:19:16.969 --> 0:19:19.729 upon termination of their employment at either the employer or 0:19:19.729 --> 0:19:23.488 their behest, will be entitled to a one-time payout of 0:19:23.489 --> 0:19:24.599