S3 #46 Why should we be vigilant when politicians talk? Brain-to-brain with Prof. Nicole Gotzner.

Introduction to Kaleidoscience

0:00Hi, and welcome to Kaleidoscience. Here you find answers about cognition that you may or may not have asked yourself.

Introduction to Kaleidoscience

0:01This episode is hosted by Imogen Huesing and Elisa Palme. So sit back, relax, and enjoy this week's episode. In the past episodes, we already talked a lot about language. Now we want to go a step further and examine communication between people. Does our conversation partner influence how we interpret language? Let's find it out together with Professor Nicole Gottsner. Professor Nicole Gottsner is currently a professor in the Cognitive Science Institute at Osnabrück University.

0:34Previously, she was an Emmy Noether group leader at the University of Potsdam and held a replacement professorship at Humboldt University in Berlin. Her current research areas include semantics, pragmatics, and psycho- and neurolinguistics. Yeah, welcome to the podcast. Thank you so much for inviting me. I really look forward to this. And as always, we play our short Get You Know game, where you get

1:05five sentence beginnings, and we will ask you to finish them as spontaneously as you can. Our first sentence is, as a kid, I always wanted to be... A language teacher. So you stayed in linguistics in the end, or language in general. Well, I found my true passion when I started my studies for becoming a teacher. So my first linguistics course, basically, I mean, as a child, I was already reading books on foreign words.

1:40And then when I was in my first linguistics course, I was like, oh, this is what I always wanted to do. And now I'm basically both, right? Because I'm also teaching in English. Yeah. That's cool. Perfect. Our next sentence. If I was an emoji, I would be... The colorful rainbow. Oh, colorful rainbow. Any specific reason or just rainbow? I think emotionally, it associates with me. And I also have like bags with rainbows.

2:15It's just the type of image I feel at home with the most. Our next sentence. My favorite thing to do on a day off is... Sauna. And workout. Okay. Right now. I think sauna is... That's the first time we had that answer, I think. At least I can't recall. Usually people like to go out in nature or spend time with friends. But I think we haven't had sauna or wellness before.

2:49That's good. We get connected to why is it called spa. Yeah. Next sentence. Right now, I'm most fascinated by... Political language. We will talk about that a bit more later on in this episode as well, I think. And our last sentence. I know it's time to call it a day when... That's something I don't know, actually. Fair enough. You already mentioned that you originally wanted to be a language teacher.

3:22Can you maybe tell us a bit more about how you came from wanting to be a language teacher to being a professor in language and pragmatics? So, I mean, basically, as a kid, I was already so fascinated by learning language, like French especially. And I always liked giving presentations at school. And so I thought, okay, I'll just, you know, start studying French and English.

3:53And then I was in my first class in linguistics, actually with Maren Huberti, who I thought of the other day. And I was in this linguistics course and I was like, this is what I was dying to learn my whole life, basically. And then I found out that there was actually, you know, an actual job you can do, just like researching language. And while I studied, I also became super fascinated by psychology, because ultimately I want

4:29to, you know, understand myself and the people around me. So I also took several courses in psychology on the side. And then for my master's, I studied linguistics and then I pursued a PhD within a priority program, basically in Berlin and Potsdam, where I was able to combine my two passions in formal linguistics and also in psycholinguistics and psychology. Then after that, I had a postdoc position at the Center for General Linguistics, where I started to work

5:06on basically implicature, also interactive communication and game theory. So this was the moment where I integrated also some of the modeling work that we're doing right now. And then I wanted to have my own project. So I applied for an Emmy Noota grant, which allowed me to have my own group. And I took that to the University of Potsdam, the cognitive science and linguistics department there.

5:38And then I saw the job advertised at Osnabrück University. And I thought this is the perfect place. It has all of the interdisciplinary research that I truly care about, a lot of topics that I'm really fascinated with. So coming here was also a moment where I was able to branch out even more into other areas of communication. For example, animal communication with Simone together in our department with my PhD students.

6:11We're doing computational work and also work on a lot of other fascinating topics like the political language topic. Something that I was also working with Nicola Compa in the philosophy department with. In today's episode, we will talk about one specific paper you published, which is, Does it matter what is said and who said it? The interpretation of Trump's and Harris statements in Republican and Democrat voter groups.

6:42Okay, before we now dive into the topic or the paper that we wanted to discuss today with you, maybe we can just roughly recap some concepts that probably will come up in the last episode. We already talked a bit more in depth about this. So just maybe a small reminder or recap. What is pragmatics? So pragmatics is basically the study of language use and interpretation and context.

7:14Should I say more or should it be brief? We can say more. So basically in pragmatics, we study a variety of communicative phenomena. So we want to develop on the one hand, formal models of how communication works. There has also been an experimental term in pragmatics. So a lot of conversations and phenomena like metaphor, figurative language, implicature. So all the type of meaning that's basically not literally encoded in a word or in an utterance can

7:50now be studied with a range of experimental paradigms and also corpora. And typically we're interested in seeing whether some aspects of the context help people understand what the intentions of the speaker was. And how does someone understand or interpret what a speaker might have meant and what they said? So just the process behind. Yeah. So this on the one hand connects to cognition and theory of mind.

8:21So we need some account also of how people are able to infer intentions. But maybe let's start from the very basics first. So a lot of what we do is inspired by the Greissian model of communication, which basically accounts for the fact that in natural language, we have this type of meaning, which you cannot directly read off from just the words that you're using.

8:51So while speaking and communicating, we also basically follow some kind of implicit contract that we should work towards some kind of joint goal. And so we have a number of expectations for how conversations should look like. So people should be informative. They should be clear because ultimately we want to have a successful information exchange. And based on this implicit contract, basically the hearer, the recipient of the language is able to infer what was

9:25truly meant in the context. So to give a concrete example, if you ask me, can you come to this episode on Wednesday? And I would say, well, I actually have some other meeting on Wednesday. Then I'm implicitly basically refusing this kind of invitation, even though literally speaking, I have not actually said that I'm not coming. And this is the interesting part about this kind of implied meaning.

9:56So on the one end, the speaker is not so accountable for what they are saying, because you can back off from what's implied. So this refusal is never literally communicated. So it's not really on record in the conversation. And on the other hand, the hearer actually has this burden of figuring out what this ambiguity is about, like what is the speaker really trying to convey here? It's just saying because you could also interpret it in a way of being, well, they just can't join on

10:34Wednesdays, but could come to an interview each other day, basically. Yeah. And that's the thing, right? So it doesn't follow that I'm not coming. But typically in conversations, we use these indirect statements to basically communicate something more. And then the interesting question is, why do we speak in this way? Like in some sense, it's inefficient. And as you indicate, like it's vulnerable to be misunderstood, right? So in a sense, it would be more efficient for me to speak in this way.

11:11But then it's also super impolite to say, no, I'm not coming. So there is some added social reward, basically, of speaking in this indirect way. And how far does it make a difference which native language a person comes from? Because I know that, for example, English is much more implicit compared to German in general. Does it make a difference? Answer me what you mean by this. So in general, pragmatics is rather universal. So to the extent that also we all have a theory of mind,

11:49you know, we all have some kind of cooperative principle that to some extent we're working with. Where you find variation is what are the possible utterances people use? And basically, how do they compete with each other? The choice that you're making in the language specifically, there would be some variation with respect to that. But the overarching cognitive and social mechanisms that allow us to communicate in this way are to some extent, at least, universal. Like this doesn't really mean that we wouldn't find also specific

12:28differences between languages. And certainly, in German, for example, it's much more acceptable to be quite direct, even for even in a case of a conflict, like even when I'm saying something that might be a bit negative, it will be more acceptable to be a bit direct here. Whereas what you find with American speakers is that even if, you know, you have a hotel star rating of one star, the American can say amazing in that context.

13:02So there's at least some, you know, production experiments, which would suggest that. So basically, amazing is completely uninformative. So you can use it in any world state. But a German would not come out saying amazing. So here, the default would rather be something like, well, it's good. Reminds me of this joke that it's edible as one of the highest compliments in Germany for food. There you also have some variation, right? That depending on the dialect you're speaking, you might be like, like some

13:40people say people in Hannover are especially like understated. They don't give you a positive compliment, basically. We had also this conversation a bit earlier where a person who is from northern Germany said that when he says, well, that's something we can do or that we could do. It's basically that he's really, he really wants to do it. But the statement to me just sounds like, well, I could do it even though I don't want to.

14:14Yeah. But yeah, and there's definitely miscommunication because of this. So because of the specific language we choose and the applications you can derive from that, certainly in cross-cultural or cross-dialectal context, there is some miscommunication arising from this. How do we then make sure that we actually get the same meaning? Yeah, that's the question is, do we get the same meaning, right? So hopefully, I think we would align in terms of the goal and know what to do next.

14:50But there's probably some, at the concept level, at least some variation with respect to the image we have in our head with respect to the situation. So, I mean, it's actually quite remarkable that people understand each other at all and even across languages. So you would think that there is quite some mismatch, I think, between the representations people have in their heads. But at least on the behavioral level, in many contexts, we seem to be doing

15:25quite well. And this is probably the result of all these implicit expectations and these working contracts that we have in communication. Is this what is meant with, is language cooperative? Or that language is cooperative? Yeah. So the cooperative principle, specifically, that was proposed by Grice, states that we should make our contributions as required by the purpose of the talk exchange at which they occur. So specifically, there is some norm for what I'm supposed to say, how much information I'm supposed to provide, and

16:02how clear I'm supposed to be. And this is something that's actually on the conversational or linguistic level. So it doesn't necessarily refer directly to goals you might have in terms of being selfish, deceiving someone or so. So the cooperative principle applies at the level of conversation, and it accounts for the fact that we have these conventions in language. So we're using language that we as a community have agreed upon, because otherwise we wouldn't understand

16:36each other. You know, if I come around and say, I want to use the word share now for something completely differently, then you wouldn't understand me. But then again, sometimes people understand a nonsense word when used in a certain context, because they kind of get the context out of, or the meaning out of the context, right? Yeah. Yeah, and this is where also the cognitive pressures come into play for the speaker, right?

17:09So if you don't know how to refer to an object, you say the thing, and then you point towards it. So you're making it easier, basically, by choosing some kind of label that can just refer to the object in the context. So cooperative in this context does not necessarily mean that you have a common goal in the conversation, but just the language. So not about, for example, if I'm lying in a conversation or deceiving my conversation

17:49partner. That is not meant with that. So, in general, there has to be some common goal to keep the conversation going and to be understood. So this is the primary concern for this cooperative principle. And then we might have these other social goals or some selfish goals to deceive the other person. But interestingly, for example, in cases where, at least in a competition sense, there's some incentive for you to deceive

18:25me, I often would still understand what you've meant based on the classic principles. So even when you've implicated something and I know that you might be deceiving me, I can still figure out that this is the implicated meaning, basically. So a good example are courtroom cross-examinations. So it's clear that we don't have an aligned goal here, but still you would be able to infer what

18:59the implicit messages are that are being sent, basically. So this is why, to some extent, we need to separate these different kind of cooperative goals. The pragmatics account Graysian principles would be mostly about the conversational goals. But then, obviously, in these other contexts where the goals are not aligned, we might still have some very interesting communicative effects. And in terms of the deception, what's also important is that we need to distinguish between what's communicated

19:38and also whether I accept the information as true. So basically, for communication to be advantages on a social and an evolutionary level, we need to be able to detect misinformation. So I have to have some mechanism. I need to be vigilant towards the source and who I'm talking to. Am I in a context where there's maybe a lot of misinformation?

20:09And then I should be able to figure out what you want to communicate, but then maybe not endorse it as true. How much of these processes are conscious? Good question. So a lot of these processes are actually quite routinized. And the models that have been proposed, it's very unlikely that at a conscious level you would go through all

20:39of these computations, like figuring out, basically, who am I talking to? What is this person conveying on a content level? What did they say? What did they not say? And so on. So those are processes that over the lifespan, basically, you automize them to some extent. How does this relate to different constellations? So when you're talking one-on-one, when you're talking in a group, or one speaker to an audience, or already relating

21:16to the experiment from the paper, when you just hear or read a statement from a person that is not live, not at the moment? Yeah, that's a super relevant question. So in principle, the standard question communication model accounts for dyadic conversations when we have two partners. So that's the classic scenario that we have two people talking to each other. Obviously, often we are in contexts where, for example, here, now there's two people I'm talking to.

21:49There's also an audience outside, which is listening to what I say. And in that sense, it's actually quite similar to what politicians do a lot. So they're kind of talking to some audience that's sitting in front of them. But at the same time, the world is watching, and they might be addressing mostly actually the people at home. So it's not very clear what the audience is in such a context.

22:19And so somehow the messages that are being sent are tailored towards achieving some kind of communicative goal. And that's not necessarily mostly addressing the person that's in front of you. So in a sense, we have to say something more about communication. We have to adapt our models to some extent when we look at the group level and also these contexts where there might be multiple audiences that we're talking to.

22:50So depending on your audience, you adjust the language you're using, the terms maybe to convey the correct information. Yeah. So, I mean, we always do that to some extent. I mean, some people more, some people less. There's also the egocentric view where in some contexts people just refer to what's easy for them without necessarily taking the listener perspective into account. But at large, we do have some mechanism of taking into account how we will be understood because otherwise it

23:25wouldn't be effective how we communicate. Now, in the political context, it's super interesting because we don't actually know like who the main target audience is. And to give you an example of how this could actually pan out differently, I'm wearing this sweater here and I wanted to show it to you as an example of a dog whistle. So this is some coded language that's very typical for political language.

23:55So basically, you're using some message or some word that's only understood by a subgroup in the audience. So in this case here, this actually reveals that I'm a huge Stranger Things fan, but you only know this if you've watched the latest episode. So for the rest of the audience, this is completely opaque that I'm sending this message now. And this is also why for the politicians, it's quite effective to communicate in this way because

24:28you can actually back off. So you can say, well, I'm just wearing this sweater because I think it's beautiful and I didn't want to convey any hidden messages here. Could you briefly describe your sweater for the people who don't, who aren't seeing the video on Instagram right now? Basically, it's a sweater that says, beam me up, this place sucks, which is also sort of a negative message. But this was the sweater that was actually worn in the latest episode a lot by one of the characters.

25:05and it kind of shows an UFO or an object and people underneath. And a woman. So it has also this kind of 70s, 60s vibe. Yeah. We might just link the t-shirt as well in the show notes so people can look at it. It's a merch. Yes. I think this already brings us to the study we wanted to talk to you about today. Could you just really shortly talk us through the setup and what was done in the study?

25:43Yeah. So the study just got published or accepted with Open Mind, which is a cognitive science journal, Diamond Open Access. and in this study I basically used a very classic task for pragmatics. So this is a task where you present people with a statement that's under informative. So saying something like the weather will be warm and there is a potential implicature here. I've used this term before. So basically implicatures are inferences that are cancelable.

26:18So you may infer that it's warm but not hot if you reason about these kinds of expressions and how they relate to the temperature scale. So this is a meaning that's not strictly encoded in the message but it's something you infer based on cooperative principles and the assumption that basically the speaker should maximize informativeness. And in the study specifically people read 70 statements that were under informative and given the setup I was interested

26:54in whether political beliefs could play any role in how you interpret these implicit messages basically. So the underlying assumption would be that when we align in terms of our beliefs it's easier to understand what's communicated implicitly. So I recruited two voter groups one group that was basically Republican and another group that was Democratic and this was pre-screened on Prolific so participants didn't know that I recruited them for this purpose and the study

27:29was run right before the presidential election last year. So at that point in time two days before the election according to the prediction both candidates Kamala Harris and Donald Trump were equally likely to win and in the experiment people saw these different statements and a further manipulation was who the speaker was so whether the sentences were uttered by Donald Trump or Kamala Harris and basically the main prediction I had was that for the Harris case of

28:05course the alignment is greater within the Democratic voter group whereas for the Trump case you have the opposite alignment with the Republican group and so I predicted a crossover interaction which is also what the results showed so basically depending on who says something you are more likely to understand what is meant implicitly and so under correctly understanding what was implicitly meant for the statement with the weather is going to be warm should the participants infer that

28:42it's going to be hot or that it's going to be warm but not hot that it's going to be warm but not hot and also notice that there is not really one correct interpretation there's always an ambiguity between the literal meaning and the implicated meaning for these kinds of statements so it's perfectly valid to say also that warm means possibly hot but the implicature meaning would be the one where you're inferring something stronger namely warm but

29:19not hot and this is where we typically need these kinds of background assumptions to arrive at that meaning so for the candidate that they were not in favor of they would rather infer the more extreme version of the stated adjective than the more precise interpretation of that I wouldn't say more extreme but the literal meaning basically that it's open whether it's hot or not hot so a way to understand this is also that you're basically spending

29:55a bit less effort thinking about what's meant if you cannot make sense of the background assumptions and I also did a further analysis in this paper where I looked at in-groups and out-groups so basically pooling the data for the groups where you have this alignment versus the cases where the out-group basically Harris is talking to the Republican group and Trump is talking to the Democratic group and you can show that compared to a baseline with no

30:32political candidates in both groups they're more likely to be understood basically so in both groups you find this boost that the implicature rates go up but this is highest for the in-group members so basically this would suggest that you're spending most effort to understand someone who is from your own group but in addition to that also knowing that you talk to an authority a political leader makes you more likely to spend some effort to think about

31:09the implicit messages that follow okay that's that's super interesting that's not what I would have expected I think what would you have expected that there isn't an effect on the like from the out group towards the other political leader like different from when the statement or if the statement was made by I don't know a random Joe or something like that so there is definitely there's some research for example by Kupa Vasa and Titrid indicating that

31:46with out group members at least if they're on the same social hierarchical level there can be some pragmatic derogation meaning that it's actually harder for you to understand what's meant and this is also connected to theory of mind abilities as they find but I think what comes into play in my experiment design is that we have these authority figures so we basically have the political leaders compared to someone like John Smith for example and you do

32:22find on the one hand that if a person is in a high hierarchy position the person in the lower position spends more effort understanding the indirect language so that's one aspect and the other aspect is also that sometimes you find basically inflated reputations so think of a cult leader so whatever the cult leader says you cannot question anymore what he's saying so in a sense you're less vigilant towards what they're saying so this might also come

32:59into play here to some extent on the one hand the hierarchy the power position but also that the reputations get inflated and for this reason you're more likely to to infer and accept the messages that are that are communicated indirectly could it also play a role that for either candidate people already probably consumed a lot of their speech like watch debates read statements online or something like that so I definitely think this has an influence especially

33:36afterwards after Trump has become the president obviously people were exposed to I mean first of all the political situation has changed but they've also been exposed to specific communication strategies and in our study project together with some students we've replicated some of the results also looking at incremental reading and we actually find that the results also shift across time to some extent to me it sounds a bit like trust also plays a role in how far

34:13we believe what a political speaker says so can you maybe tell us a bit about the possible role trust could have on that yeah this is precisely also what the framework that I'm using here would argue so basically this epistemic vigilance view by Sperber and colleagues from 2010 so they would argue that we have these cognitive mechanisms which allow us to evaluate the source and the content and so we tend to award some stance of trust

34:50typically which also aligns with this cooperative assumption because this is just effective for communication to work but then in certain contexts we know that we need to be vigilant and not accept all of the information that comes through so trust is certainly like one of the key mechanisms that would also affect these kinds of conversational dynamics and what does this mean for us if we now interact with political statements from figures that align with our beliefs

35:28or the opposite completely misalign with our beliefs how should or is there anything we should maybe consider when interacting with statements yeah I think in either case no matter if you align or if you don't align it's very important to be vigilant basically and that doesn't only apply to politicians but also to other contexts like when you're in contexts where you know there's a lot of misinformation in general it's a good strategy to you know double

36:05check what information am I being presented with so this basically media competence is also like really important for society at large and more specifically for the political case I think we should factor in that to some extent in political communication there's a payoff for politicians to be vague or indirect because it allows them also to pursue their own goals without going on record about something that they might regret later essentially that makes sense and also just

36:42allows to step back and say well I didn't mean it that way yeah I mean you find that really on record a lot in these conversations and we're spending all this time you know going through examples of like I said it I didn't say it I didn't mean it and so on but especially when it comes to certain aspects you are accountable so you're definitely less accountable for implicatures and indirect language so you can show this

37:19because I can retract later so coming back to our refusal example I can say well I have this other meeting of course I'm happy to come there would not be any contradiction following up in this way but then there are other cases where I'm using something that's offensive and there I cannot easily retract and you have to take the responsibility for what you say in this case in this whole research which role does cognition play so

37:56cognition plays a role in terms of on the one hand the cognitive biases we might have when we engage with different political figures but also in terms of the socio cognitive machinery that we have allowing us to communicate so basically this whole epistemic vigilance view relies on the idea that we have a theory of mind and we're able to judge what information comes in and also take into account what we know about the speaker so those

38:33are all cognitive abilities and for the moment we've looked at basically offline data judgments and final interpretation but we're also doing a range of studies looking at the online incremental interpretation of language and there was some interesting work for example by van Berkum and colleagues who showed that basically immediately once we hear a word that's inconsistent with our moral values we see a sign of that in the brain activity so we tend to integrate and update

39:11our beliefs with the information that's coming in quite rapidly in this case the study looked at basically moral statements which is different from what I did here because I was interested in the implied language aspect so here in our case we would expect for such effects to be rather late and downstream not necessarily showing up immediately incrementally but we can certainly find out a lot of more interesting results on how cognitive biases factor into the results

39:48and other processes related to incremental language yeah and what is the one thing that our listeners should take home from this episode today it not only matters what is said but who says it great that's a really precise short answer finish strong and with that said thank you so much for joining us today and talking to us thank you for inviting you also yeah this was kaleidoscience we hope that you enjoyed this episode and we would

40:25love to have your feedback you can rate our podcast and give us feedback on our instagram account have a great week and you'll hear from us again in two weeks this episode was hosted by imogenhüsing and elisa palmer produced by imogenhüsing clara kühne sophie kühne söhnke lülf and elisa palmer the music is from jan-luka schröder and the logo is from annika richter overเร do on on on is we on on