Full Transcript for How Should We Talk To and About AI?

So we used to write commands in a terminal, right? It would be like literally be called the command line. And now there'll be questions and there'll be like these like polite kind of ⁓ phrasings and they're using articles and pronouns in a way like is very enforced by the output that comes from these chat bots, right? So the the someone re recently sent me an example where They had asked for like what are good ways to rent a car in this travel destination I'm going to. And then the output was something along the lines of like I don't have much experience renting cars. Which ⁓ this is unsurprising. But the fact that it's saying I is just weird, right? So people will lean into this 'cause that's how we communicate. speaker-0: Yeah. Okay, excellent. We'll definitely come back to it. But I want I want to bring in Webb. And so my question for you is how do the ways that people talk to and about AI resemble the other ways that we've talked about non human creatures? So I'm thinking about inanimate objects. I'm also thinking about forces throughout history, like natural forces, and then also non human animals. speaker-2: Yeah. So much of the hype about and so much of the anxiety about ⁓ AI has been basing the idea that would that we're encountering something that's so totally and radically other than anything we've ever encountered before. Part of what I try to do is say, well, of course there is something radically new going on here. No one would deny that, but to overemphasize that is both to misunderstand something very important about it and also I w I might add to give far too much credit to the corporations that run it. But that's another story. It's not another story, but it's a story we'll get to. ⁓ We in fact, humans have an extremely long history of doing things like this. And some examples I give in the book come both from domesticated animals, but perhaps even more interestingly from victims of hunters and sacrificers. So it's very common in societies where people Primarily lived by hunting, such as people in the Yukon and Siberia traditionally, to interact with their prey as if they were human or human-like, sometimes as reincarnations of ancestors, sometimes as something that's not human, but is near enough to being a human that you need to deal with it in properly human ways. And one of the things that follows from this is. That you need to understand the ethical implications of what you're doing by killing it. So no one who hunts animals for a living, and I'm thinking about like bear hunters in the Yukon, for example, ⁓ they are under no illusions about what a bear is. They know that bears are not human, they know that bears are dangerous, can be dangerous, but they also know that bears are animate beings who you are going to kill. And you need to kill in order to survive. So you might say that here, anthropomorphizing them through acts of addressing them in the second person, you, grandfather bear, ⁓ you need to understand that that I have to do this to you. And I want you to understand that, is a way of taking, in some sense, taking responsibility for your own actions. One way of thinking about it is to say that rather than seeing The killing of the animal as a neutral act in a morally neutral universe, you're essentially saying what I do matters because what happens in the universe matters because it's full of ethical beings of whom humans are just one. And it's so it's a way of sort of saturating the world with with sort of ethical attentiveness. speaker-0: Well this is fascinating. I'm wondering if we could sit with this question of whether the anthropomorphizing or anthropomorphization that's happening, the extent to which it's new and the extent to which it's continuous with past practices. speaker-2: Yeah. So well, let me give you an example that that involves robots. That so they're not LLMs, they're not AI, but this example from my book, which is ⁓ the popularity in Japan of robot pet dogs. And you and I will agree, ⁓ I assume Nana will agree that that robot pet dogs are not animate beings. and apparently anthropologists have reported that there are they're very popular in Japan, and there are there are Buddhist temples that will undertake Essentially mortuary rituals for robot pet dogs that have ceased to function. And you might then say then, well, the owners of those dogs are operating under an illusion. They're just mistaken about metaphysical facts. And it's possible that they are mistaken, but there's another possibility, which is their relationship to those dogs is an as-if kind of relationship, that it is. They play with the possibility they're playing along an ontological borderline between what is real and not real and moving back and forth across it. And I think we do this all the time. I think we do this when we develop relationships with characters in novels or movies. Yeah. Which are an awful lot like real relationships. You may even feel bereft when you finish a long novel and you say, I miss them right. But you're not under an illusion that they're the same as, you know, your next door neighbor or your partner, your children. speaker-0: Yeah. So I I I want to dig into all that, but first I just want to let Nana have the opportunity to respond as well. And so just just bringing us back to this this question, what I'm wondering about is when we think about the fact that the LLM is not actually a thinking being who's experiencing reality, and the bear is a thinking being who's experiencing reality. Is that helpful for thinking about continuities and discontinuities with past practices of anthropomorphization? speaker-1: I think there's a huge difference in that the LLMs are controlled by corporations and can be changed by design, right? There's some human choices behind all of what goes on there, whereas the bear is its own entity. And even if ⁓ we at some point ⁓ decide that some LM has now developed consciousness, yada yada, we can still pull the plug,

right? So there's a whole different level ⁓ different level of agency that the human still has and that I think makes these two entities not very comparable, especially because it's kind of like imbuing this agency within a model that actually lies with humans, which I think is like potentially really dangerous and removes the accountability of the people that design these tools. speaker-2: So I so I I'd like to if I could, I'd want to jump in and introduce say I say I'd like to introduce another term of art, which is not anthropomorphism, because I think that there's continuity in anthropomorphism. But the distinction and I agree with Nana, and I think the distinction here is something that that ⁓ we could call fetishism. And by fetishism, you know, I don't mean whips and bonds and stuff like that, but what I mean is the displacement of agency from something that you have made onto the thing that has been made. And so essentially the danger here is not recognizing the human agency behind the design of LLMs. And furthermore, to by virtue of the way in which speaker-0: Mm. speaker-2: The anthropomorphization of the LLMs tends to lead to the attribution of superhuman powers to the LLM. Essentially what you're doing is you're without even realizing it, you're granting those powers to the corporations. So they're simultaneously denying the human role in the design of the LLM and inadvertently attributing the power of the LLM to the designer. It's kind of a funny kind of a loop there. speaker-1: Mm-hmm. speaker-0: That's really interesting. speaker-2: And so we're grant we're granting we're granting this sort of massive authority to these corporations without even realizing it. speaker-0: Hmm. So just let me make sure I understand. So it's like we treat the LLM as if it has these godlike powers. It's this mysterious entity that we interact with. We ask it whatever questions we want, we get these answers. And I think that that's that goes back to what you were saying saying earlier, Nana, about the replacement of Google, where now you just ask ChatGPT your Claude, your question, rather than Googling it. So we we interact with it in this way that almost is like we're interacting with an Oracle. But then we're not thinking about human designers. We're not thinking about an individual human exercising agency when we're interacting with it. But these large corporations are actually designing these systems. And so then we're putting them in this godlike position. Is is that a fair summary? Yeah. speaker-2: Yeah. So we're making in some sense the oracle is is controlled by the corporation. You so think of the Wizard of Oz, you know, ignore the man behind that curtain, right? There's just like there is this oracle, but behind the Oracle is the corporation. So the authority we grant to the Oracle essentially becomes is absorbed by the corporation without us even realizing it. Yeah. speaker-0: This is exactly where I was hoping to get to with this question, because I think what what I'm seeing in these answers that we're coming up with is that the continuity is the fact that we treat it as if it was an oracle. And this is an ancient practice that has, you know, emerged in cultures across the entire world simultaneously and and without any interaction between them in some cases. But it's misleading in the sense that when you're interacting with an oracle, you're not actually treating a corporation as if it's a god. There's some there's some other explanation for it. But in this case, we're actually elevating anthropic even though we don't know it, or open AI into this godlike position. speaker-2: Right. Although, you know, the Delphic Oracle was speaking on behalf of the god Apollo. But what what they both have in common is that that their authority also draws on the fact that their workings are enigmatic and inscrutable. You can't see inside the black box. And therefore, it invites us to think they must be in touch with some absolutely transcendental source. speaker-0: Yeah, well that's fascinating. So my question is I think we can all see a role for when you're trying to understand somebody from another culture, framing it in terms of their own language. And when I'm thinking about the Japanese robot dog case, the way that I'm tempted to do that is to ask a couple of questions. One question is: does the dog actually have conscious experiences? And that's not a question that I Have access to directly or anybody else does. It's just it it's a question that the dog would have access to if it was conscious. And if it's not conscious, it's just not conscious. But if I'm interacting with somebody who behaves in that way, I don't want to presuppose that I understand the value of their way of treating the dog as if it was conscious. And I don't want to tell them that they're mistaken for engaging in that social practice because it might have some meaning and importance and social role that I don't even understand. And so the best way for me to actually understand this person might be for me to act as if all these ontological claims were true, even if they're making claims about the dog that are actually false because the dog itself does not have a perspective. speaker-1: I think there's there different things are valuable at different points, right? So is the point that you want to understand those people or is the point that you want to save these people from themselves? In different occasions this like different things can be useful, right? So there's that pet seals paro. ⁓ I believe that they gave to elderly in nursing homes and it was actually really beneficial for them. So we can I think there no one would deny them this pleasure. speaker-2: Can you explain that, please? speaker-1: There used to be this like robot seal that was like a s probably similar to these robot dogs, but it was fluffy and it was warm. And they would give these robot seals to elderly in nursing homes to comfort them like a

comfort animal but without the responsibilities of feeding and walking, etc. And I think the general research and experience has just said that this brought these people a lot of pleasure. And here I think it's very like there's a a pretty easy benefit value to these people. using or petting or holding this seal, right? So in that context, I don't think there's any point in going around telling them this isn't actually an animal or 'cause it's like it it brings them benefit. Now, there are some harmful misunderstandings in the world, right? Such as when this man who was using chat GPT to to ask for an alternative to sodium chloride in his diet, because he was trying to protect his blood pressure, and he got The suggestion of using sodium bromide, I believe, and then got a case of bromide poisoning. It's harmful, right? And we need to there protect the man from himself who is trusting in a harmful way to himself this output, right? So I hope that's that answers or at least addresses your question. I think sometimes there's an advantage to actually telling like your understanding of the world isn't doesn't match what a lot of the rest of us understand, ⁓ or it's harmful to you or others. And then sometimes it just seems less harmful, right? And who am I to say that someone can't perceive an entity in some what I believe to be dead object? speaker-0: Yeah, that's fascinating. I just want to make I'm getting it. So the idea is like the seal is not actually conscious, but it's okay to it's not even okay. It's like it could be actually beneficial to act as if it's conscious. And in fact, I could be wronging somebody by going in there and trying to disabuse them of the thought that it's conscious. And there's other cases where it might but where it could actually be beneficial. So you tell the person, hey, chat GPT is actually not conscious. ⁓ you shouldn't trust it as a medical authority, especially when you're thinking about how you want to make These dietary changes. That's the idea. speaker-1: Exactly. That's why we were in our research, we were looking for like how does anthropomorphism or anthropomorphization actually look like when we talk about these tools, right? So how does this happen by use ⁓ like in terms of the language that we use? And there are different categories of anthropomorphization. One of them could even be are we doing it ironically or not, right? Like when I say my car seems to really be having a bad day. This is clearly ironic anthropomorphism where I don't think the car is human. ⁓ and I think this category is like less harmful, right? 'Cause we're clearly flagging that this is a metaphor and I'm making a joke or an ironic statement. And then there are all of these linguistic traces or patterns or uses of language where it becomes a lot more fuzzy. Like that where the difference is way less clear cut and this might potentially be harmful. speaker-0: Got it. So your concern isn't just with misleading uses of language or uses of language that aren't literal. It's with times when we do that and it actually harms us. speaker-1: It's just yeah, I'm pretty pragmatic and like I don't see any reason to like if flat earthers want to be flat earthers and they're not harming people, go for it. Or if people want to believe in different gods and they're not harming anyone or themselves, go for it. But I do think that when this anthropomorphization obfuscates the responsibility of big corporations and it does harm people, there's a big reason for us to have a critical discussion. speaker-0: Excellent, okay, good, thank you. Webb, sorry, ⁓ go for it. I know that you wanted to jump in. speaker-2: I think I mean I I come I agree. And I th I think the though I would I would make this less a question of ontology, like what it what's what what they really are, and more a question of the nature of the of the interactions and and what follows from that. So where I would push so so I also know I know your work with with Emily Vender and what you at least in the work that I've read, you focus on how we talk about these things. And you know, I am a big fan of the people like George Lakoff who have argued that we are affected by the varied metaphors in our language. But I think what is ultimately more consequential, I think, is is the metaphors that are built into the practices rather than the referential language. So it's actually well, first of all, reforming vocabulary is a lot harder than you might think, as as all you know, his the history of linguistic reform shows. Really hard to change people's vocabulary. But even if you could, it will how people talk about devices will not, I think, at the end of the day, be as powerful as how they interact with them. So there you might say that there is a metaphor built not just into the language, the referential language, the language, the ways we talk about them. There's a pragmatic metaphor built into the ways we deal with them. And That's where things really become very stubborn and very consequential. So I could change the way I talk about it, but if I still interact with it as if it were an authority, now we're in now we're in trouble. speaker-1: I agree completely. The thing is, we don't have much control over how the output of these LLMs look, right? It's corporations. So someone introduced me ⁓ unkindly to the concept of ⁓ anthropometicism, which as if anthropomorphization wasn't annoying enough to say. ⁓ but anthropomemeticism is the way that these tools mimic humans, right? And that's a whole different area, but one that we don't have so much control over. Whereas I think what we were trying to do in our work was to identify, well, at least we the what we could do as researchers, hopefully journalists, ⁓ is to kind of flag that we're saying something that isn't true when we're using that kind of metaphor 'cause I also say chatbots, right? Even though I love the term conversation simulator. ⁓ speaker-2: Yeah, it just comes tripping off

the tongue, right? speaker-0: Right, yeah. speaker-1: Yeah, it's very catchy. No, but we're actually like what we want to look into next is whether helping people flag these metaphors or like be aware of that language, whether that kind of enhances their critical evaluation of the output or that actually changes their thinking and behavior. Because you I think you're completely right. We can this is a dead battle if the output continues to look like human output. speaker-0: I I wanna ⁓ go back a little bit just to make sure the audience understands what we're talking about when we're just thinking about the difference between referential language and then the social practices that are manifesting how we think about AI. And so I I'm gonna give an example and I and I just wanna see if if this resonates with with the two of you for how you're thinking about that. So when you, Nana, are criticized or you're thinking about errors in how people are using language to describe AI. One thing you think about is names and pronouns. And so you might interact with a chatbot and ask a question like who is right? But rather than asking who is right, what you'd want to have them say is, is the machine output correct? Or rather than saying something like, they produce results, like the machine produces results, you'd want to say the team used it to produce results. And I think the idea there is that. We can intervene on language. We might, you know, we we can like that's something that we can actually consciously choose to change. I think, Webb, you've suggested that it's not as easy as you think. ⁓ I've been active in like animal rights activism for about fifteen years now. And I've also seen that it's hard to get people to stop saying things like, ⁓ I don't know, kill two birds with one stone. I want them to say like I want them to say feed two birds with one stone. And I've been trying to do that for years. Or things like don't beat a dead horse. I try to get them to say, Don't feed a fed horse. Like that's a little better. I would speaker-1: Have to absorb that bird thing. If you we can do a swap. I pro promise to always say feed two birds with one scone if you promise to always say conversation simulator. speaker-0: ⁓ that's that's a tough bargain. ⁓ okay, I'll do it for the I'll I'll do it for like for the animals, like if that's what it takes. One more ⁓ and so I think I think it's it's hard it's easier said than done to get people to actually change the language, but at least that's something we can point to. But you but they might change the language and then still think of it in the same way. So they might still ask, is the machine output correct? And then still think of the machine as this being who is right or not right. And I feel like that that's the crux of what the two of you were just talking about is that at least for the for the first part of it. And I guess it sounds to me like you're both on board with the idea that there are some ways that we talk to AI that can harm us. And that it would be good if we stopped doing those things. I think like the example of of the sodium one, ⁓ that's a really clear example where I think you're both on the same page that it is a mistake for him to interact with AI. speaker-1: Yeah, for sure. speaker-0: In the way that he did, it led to health problems. And my question is, is the issue that he was treating it as if it was human like, he treated it like he anthropomorphized it, or is the issue that he anthropomorphized it in a dangerous way? And that's not a leading question. I'm actually really curious about where the two of you would locate the mistake that he's making. speaker-2: I think he's granting an authority that actually doesn't in this case doesn't depend on anthropomorphism. ⁓ so I don't know if that's a good case of anthropomorphism. It is a good case of the dangers of AI. Yeah. speaker-1: So yeah, no, I I agree that this might not be the like what we're trying to look at here ⁓ is what are the dangers of anthropomorphizing actually. And of course this is not a simple, clear cut kind of question. What we've been looking into a lot in our research is whether A, does anthropomorphizing and anthropomemeticism contribute to misplaced trust? And they're like so it becomes like a modifying a link between the interaction and then a a harmful outcome because it kind of exaggerates or even creates a trust that is unwarranted. ⁓ and the other one is whether presenting these tools in this anthropomorphic way makes us evaluate their output less critically or evaluate the functionality of these tools ⁓ less critically than is than we maybe should. So we might accidentally accidentally, we might Use them for purposes where they're not appropriate. speaker-0: Mm-hmm. speaker-2: So d you might say that in in fact the question of trust is anthropomorphism might encourage us to be a little more skeptical of the answers they give, because you don't believe everything another person tells you necessarily, whereas you might believe something that a godlike machine tells you. speaker-1: Yeah, I think there's like an interesting I'm actually looking into in my like a parallel kind of work trajectory, whether like what are the different potentially harmful biases and like kind of effects that happen. One of them is like the automation bias where we assume that because these technologies are built on big data, they must therefore be more correct or objective than anything that human could write. ⁓ and then I think there's a intr I have an interest or I think an interesting question of like Do we assume that the output is based on big data, or do we confuse it with wisdom of the crowd? Right? Because even for people that know that this is based on tons of fora and news articles and stuff, they might confuse it with wisdom of the crowd, think that this wisdom of everyone is imbued in the answer. None of these perceptions are true, ⁓ because the statistical

prediction of language is very different from either of those two, right? But that's a different way of viewing it. Are we are we overtrusting maybe because we think it's based on big Objective data. speaker-2: Think that's that's more more dangerous than the in this case anthropomorphism, which is the the the the appearance of objectivity and omniscience. But if if I could take the go back to anthropomorphism and then flip it, and that is to say, and here this is an idea, if you know the work of Lucy Suchman, ⁓ and I think Sherry Turkle and there's an inversion of anthropomorphism, and that's the danger that we will start to think of ourselves as being like computers. or as being like AI. And I think there's a real danger there that people will start to identify and this gets me to my theory about w why smart people think that AI is about to be conscious. And it has to do with with a well, I don't know if we want to go there, but speaker-0: Mm. speaker-1: Go for it. speaker-2: If you think that what's most important about humans is the mind. And if you think about what the mind really is about is rationality. And if you think that rationality is really about calculation. And if you think that all takes place in the brain. And if you think that what's going on in the brain is chemical circuits moving around neural networks, then if you can reproduce a neural network in a silicon-based rather than carbon-based system, you can then take all those steps backwards, back outwards again and say, ⁓ we've got a mind. But it it the whole thing turns on a this highly reductive model of what a mind is. speaker-0: That's really interesting. So is it it's like we we have this model of what a mind is, it has to reduce to neural networks and other mechanistic features. And if you think that so is is the idea that it's wishful thinking, because if AI was conscious, it would show that our models of consciousness, the mechanistic models are are are correct. speaker-2: Well, I think it's you know, it would s there's some sense, ⁓ even in good science, you often end up finding what you're looking for. Right. I mean, you know, the the so that you what Maslow's hammer is. It's like if you ha if what you've got is a hammer, then everything looks like a nail to you. And I think if your model of consciousness is this highly re highly reductive notion and then you can reproduce in computers, it's gonna reinforce the idea that, well, that must be what consciousness is. And it's you end up with this very sort of narrow view of what a human being is. speaker-0: ⁓ that's interesting. ⁓ I think it's a great tool as long as you're maintaining human interaction. I I think over relying on this can you you don't want to be speaking and conversing and interacting with AI so often that it you bleed bad tendencies into your human interactions as well. With humans this is a two way conversation. You have to take into consideration the other person's conscious mind and that you know, be ⁓ empathetic towards that. Not saying we shouldn't be empathetic towards the robots, but you know, there's clearly that distinguishment between them and us. There's one question that I I also wanted to to to explore. And so towards the very beginning, Nana, you were talking about the ways that people are polite to AI. So I find myself inter when I'm interacting with Claude, so Lims, I will I'll say thank you. And there's no reason to say thank you in terms of its intrinsic properties. It doesn't experience that. It's I'm maybe I'm using a couple of it's basically like I'm I'm wasting a little bit of water, is is what I'm doing. And I think that that is one way that we anthropomorphize it. Another thing that happens is people get really frustrated at it and they'll say things to AI chatbots that they would never say to a human. And when I think about the parts of Immanuel Kant's philosophy that we tend to emphasize in like introductory philosophy classes, one thing that a lot of students are going to come away from ⁓ those classes with this idea of We have duties to non-human animals that are grounded ultimately in our duties to humans. And so the idea is that when we interact with an animal in a way that harms them, we are training ourselves to be more open to harming humans. And this is not just a Kantian point. Like I consider myself an act utilitarian and I'm really concerned with what kinds of habits I'm inculcating within myself because I think that those habits are going to determine my future behaviors. If I was a virtue ethicist, I would even have a more direct route to that because I would think that the habits of thinking, feeling, and behaving that I am inculcating within myself are just directly the most important, morally relevant factors. And so one one one other question I just wanted to get both of your insights on is this this Kantian point about about shaping ourselves. When people interact with AI in ways that they would never interact with a human because they're just nasty to the chat bots in ways that would be nest if they were talking to an actual human. Is that something that ⁓ either of you are concerned with? speaker-2: You know, I once I first encountered this when I saw someone who had known for many, many, many years who was a very polite person, and I heard her suddenly bark a command at Siri. Like, Siri, play Frank Sinatra in this like imperious way that I had never heard from her before. And I and I just thought, you know, she's learning how to give orders as if she were ordering a slave around. And like what f what might follow from that? I think the I guess I'm sounding more like a virtue ethicist in this case, but maybe that's That's ⁓ speaker-0: Yeah, well we can all get on board with I mean it it's it's it's shocking. I think we can all relate to that case to some extent. Mm-hmm. speaker-1: sure I'm trying to remember I'm sure I saw a study about someone that interacted with service or support workers or a

chat bot where they were behaving really rude and where they showed that this carried over to our interactions with others. But this is me like presenting it really badly because I can't actually remember this study. But I I'm quite convinced that I've seen some studies of how this actually does affect our behavior towards others post the interaction. speaker-2: Yeah. speaker-0: Yeah. I w I wouldn't be surprised. I mean, when you look at like there was a debate within psychology for a while about whether the catharsis theory is true, where like you you you express some anger and then you can get out of your system and then you're nicer. And it just like the balance of evidence goes against it, where you're nasty to people, you do all this aggression, and then later on you're more likely to do that to other humans. And so I would be ⁓ shocked if the trend was reversed for AI. Sure. And I think what's what's interesting to me about this is speaker-2: Is it? speaker-0: All three of us are concerned with these pragmatic questions about what is the actual impact of the ways that we talk to and about AI? What's the impact on ourselves? What's the impact on our ability to think clearly? Like what what's the impact on our social practices and on the social practices that characterize society more broadly? And one possibility is that the things that we dismiss as mistakes when people like like at like ⁓ Mistakes in reasoning of treating AI too human in terms of being polite to it using pronouns might be something that we actually want to inculcate in our practices in order to develop the right traits towards treating other humans better. speaker-2: So I you know, I I was in a conversation with a a colleague who said, you know, he he had a he had a son who was ⁓ very young and he needed help going to sleep, he was a little bit afraid of the dark and so they gave him an Alexa or something whom he could when he was ready, he could say, Turn out the lights. And he overheard his son, you know, yeah, saying, W doing the please and thank you thing. And at first this guy was a philosopher. So his first reaction was ontological mistake. He doesn't understand it as just a device. And it it took him a while to realize, like, yeah, that's how I want my son to grow up, is is, you know, having this sense of of ⁓ politeness to others. And you start there. You know, little kids start, they learn a lot about other people from their remember, imaginary playmates, right? speaker-0: Right, right. speaker-2: Again, characters in novels, ⁓ characters on the stage, Greek tragedy are these are all forms of sort of ethical inculcation that depend on your ability to develop these ⁓ anthropomorphic relations with things that are not quite what you're making them out to be. speaker-1: May I ask why Alexa not just a light switch? speaker-2: I think with the idea being that that he was in bed and snuggled down and the blankets over him and that he wasn't gonna get out of bed and trot

over to the light switch 'cause speaker-1: I used to just have a like a lamp on my wall. I just flicked this. speaker-2: ⁓ well maybe they didn't have a lamp on the wall. We really didn't get into this. speaker-0: Well, Nana, I'd love to hear you say more because I was watching your face throughout that story and you went on such a journey, like you were horrified by some of the things you you heard. I I think the audience can't see it. ⁓ what's your take on a story like that? speaker-1: well, I might have feared that this would go into a direction where like the kid could then interact with this with this device to so as to feel less alone. And then it was like it was just to turn off the light and I'm like, what was wrong with a good old switch? ⁓ but maybe there is more to it. Maybe there were stories, like he could control his like night type stories or something. and as someone who's worked a bit with quite a bit with ⁓ learning technologies, I am particularly worried about children and teenagers, ⁓ because they're in a like their mind is in a whole different place, right? So one thing is w like I I I'm worried less about the de-skilling of adults, experts, subject experts that are using tools as a support than I am of the people who are supposed to develop certain abilities and capacities. And especially this like to go back to that ⁓ the isolation that you spoke about, like I thing it's hard work to learn to integrate socially. It is dead hard work, right? Like everyone who's ever started a new job or school knows that that this is difficult. But it becomes less difficult as we get older, hopefully, ⁓ because we learn that some of this worry we have about other people not liking us, et cetera, is is actually not true, right? Or even if it is, it might not speaker-0: Yeah. ⁓ speaker-2: Yeah. speaker-1: Or that. and that. ⁓ yes. So but I think there's like we have to learn that kind of social stamina and resilience. And I worry so much about kids that don't learn that resilience because they're overprotected and because they are so these tools are made to agree with us, right? Like they're not made to contradict us. They're not made to create any of this resilience. And I'm gonna make a weird parallel here, but there is I've recently read about studies, not just talk, it was actual studies that showed that men particularly who used a l lot of pornography were less likely to see their partners as attractive because they were expecting different things. And this like is kind of like it makes sense, right? If you're seeing a lot of like you can just streamline everything, you get less understanding of and like acceptance of like life's weirdnesses and people's weirdnesses and f fat flaps and what do I know? So here I would worry a lot about these kids if they grow up with these interactions that we're all gonna be super polite to us and we're all gonna agree with each other, that they don't develop that kind of social stamina that is required to integrate into new environments. speaker-0: Hmm. Well, let me ask yeah, so let me ask you a question. I Nana, you're talking about the ways that learning technology is interacting with young minds. And one thing that I wanted to draw a connection to is the difference between foisting all this on people right now without centuries of preparation and the ways that the communities that you study at Webb have developed like rituals and practices over hundreds or thousands of years that help them navigate it. Where when I read the stories that you tell about hunters engaging with their ⁓ prey or or the the the ones whom whom they would call sacrificers, or the ways that even though even the ways that like the s the ⁓ the the blessing on the ⁓ robot dogs are integrated within this existing worldview, it's still a way of starting with something that people understand and that gives them a sense of meaning for these practices and then helping them navigate it. Here we're just throwing stuff at kids that we don't understand and for which they have no broader meaning making structure that gives them anything close to the level of advice that people are being given in other cultures when they're engaging in anthropomorphizing or deification. And so I'm curious if that's another way of thinking about the extent to which it's new and the extent to which it's not new. speaker-2: That's a great question. And the honest truth is I I I I'm of two minds about that. Because on the one hand, I worry a lot about the speed of which which this is all happening. And I think you're you know the concern expressed in your question, I think, is is really well taken, which is ⁓ w can we adapt fast enough to sort of get control over this? That's on the one hand. On the other hand, I also think that people have this extraordinary that that w we need to give people credit for their creativity and their capacity to surprise us. And you know, when I think back to previous moral panics about television, about radio, ⁓ you know, that in hindsight, that well, they certainly had consequences, but the people also were able to speaker-0: Mm. speaker-2: take command of their lives in ways that were not you couldn't have predicted at the time. So I'd like to have that faith. How much weight I want to put on that faith is ⁓ I I'm not so sure. So I'm I'm I'm conflicted about that. I mean there there was moral panic about ⁓ when telephones first came in because, you know, dad didn't know who his daughter was talking to. speaker-0: Mm. Well, and if if we even go early, like like Socrates has some moral panic around the written word. And by the way. ⁓ speaker-2: Yeah, because I work with people who don't ⁓ were mostly illiterate. So Pl Soc Plato or Socrates was convinced that it would destroy memory and having lived with people who didn't read and write, I you know what? He was right. They have phenomenal memories. That just astonished me. speaker-0: That is fascinating for two reasons. The first is I think a lot of people will bring up this thing that you see in in in Plato where he's talk where he's talking about ⁓ being concerned with the written word destroying memory. And they kind of dismiss that as an early fear that ended up like being proven wrong. So it's interesting to me to learn that he was actually right about about that particular claim. And I think the other thing about this is that I'm still really glad that I have the written word. And I couldn't have read like your papers or your books if it wasn't for that. But it's like I'm glad because of the trade offs that I'm happy making. speaker-2: So g guess what? I think that having the written word makes us a kind of cognitive cyborg. speaker-0: ⁓ that's interesting. Yeah. speaker-2: Because our minds are at some level inseparable from reading and writing, which are technologies external to us, but also thoroughly internal to us. So I I think that makes us can you know, cognitive cyberworks. Yeah. And so we've offloaded the stuff that used to be done by memory, and it comes with a cost, but it also has, you know, there's huge benefits. speaker-0: Yeah. And that so that's something I think is is the kind of the second application that I find really interesting about what you were saying about ⁓ the memory benefits of not having a w a written word. When we're thinking about the harms and benefits of the ways that we talk to and about AI, it's not simply about noticing some bad making features. It's about figuring out which package of benefits and harms we would like to choose. I think that people at Tipley are not very thoughtful about making those choices. And they're just, you know, moving along and adopting the ways of thinking and talking to and about AI that they see other people doing. And what I'm hopeful for is that something like this podcast can let people stop and think more carefully about how they would like to talk to and about AI so that they can figure out the right balance of trade-offs of harms and benefits. for the lives that they're trying to live. All right. Well, that was a really fantastic conversation. Thank you so much, Nana. Thank you so much, Webb. It was wonderful to get to know both of you. I've enjoyed engaging with your work and I've gotten to enjoy engaging with the two of you on the air. speaker-1: Thank you very much. speaker-2: Thank you. It's a real pleasure. speaker-3: This series is a project of the Center for Ethics and Human Values at the Ohio State University. It was produced by Jennifer McCord and Carrie Welsh. speaker-0: Nana's article and Webb's book, the pieces that inspired this episode, are both linked in the show notes, along with other research referenced in the conversation. speaker-3: As always, we look forward to your voice in our ongoing conversations. We'll see you next episode.