How Algorithms Are Rewiring the Way We Talk - The Assignment with Audie Cornish - Podcast on CNN Podcasts

CNN

CNN Podcasts

5 Good Things: Mahjong: It’s Not Just Your Grandma’s Game Anymore
5 Things
Listen to
CNN 5 Things
Sat, Sep 13
New Episodes
How To Listen
On your computer On your mobile device Smart speakers
Explore CNN
US World Politics Business
podcast

The Assignment with Audie Cornish

Every Thursday on The Assignment, host Audie Cornish explores the animating forces of this extraordinary American political moment. It’s not about the horse race, it’s about the larger cultural ideas driving the conversation: the role of online influencers on the electorate, the intersection of pop culture and politics, and discussions with primary voices and thinkers who are shaping the political conversation.

Back to episodes list

How Algorithms Are Rewiring the Way We Talk
The Assignment with Audie Cornish
Sep 11, 2025

Social media doesn’t just decide what you see — it shapes how you speak. Audie talks to linguist, author and educational influencer, Adam Aleksic to explore how the way we communicate is mutating in the age of TikTok, Trump, and ChatGPT. 

Adam’s book is called, “Algospeak: How Social Media Is Transforming the Future of Language.” 

-- 

This episode was Produced by Lori Galarreta 

Senior Producer: Matt Martinez  
Technical Director: Dan Dzula   
Executive Producer:  Steve Lickteig 

Episode Transcript
Audie Cornish
00:00:00
I'm Audie Cornish and this is The Assignment. The algorithms that run the internet are changing the way we communicate in real life. For example, take the idea of an accent. I'm from Boston, which I know you can't exactly hear in my voice, but that slow, calm NPR voice, I've got that nailed.
Audie Cornish
00:00:22
Adam Aleksic, thank you so much for being with us. I see you've written a book called Algospeak, and then that would be the, like, public media way of doing it.
Audie Cornish
00:00:31
'This doesn't really work, like, at all when I'm on planet social media. Where the native tongue is memes and jokes and ever-evolving slang, and most importantly, speech.
Adam Aleksic
00:00:42
But if I'm on social media, I'm stressing every word to keep you watching my video. I'm up talking. I'm making sure you keep watching everything I say.
Audie Cornish
00:00:51
'These new online accents are shaped by algorithms, the personal recommendation engines that feed each one of us a different stream of content based on when we click and when we scroll. Like linguists are finding that we're using that weird AI vocabulary everywhere. So you might hear someone saying something like, our language model just keeps iterating, prompting new output. So everything's getting optimized into one big collab. And even worse, the out-of-context generational slang. As in, low-key, we're just vibing and talking like actual memes out loud now. Internet culture won. No cap. So today, we are talking about why this is happening with a linguist, a language detective. Stay with us.
Audie Cornish
00:01:43
Adam Aleksic is a linguist, an educational influencer, and the author of }Algospeak: How Social Media is Transforming the Future of Language."
Adam Aleksic
00:01:54
What developed under the surveillance state is now a product of surveillance capitalism, which I guess says something about how America does stuff. But it's now a tool for these platforms, the organization, the way they distribute content and classify content is all power for them.
Audie Cornish
00:02:10
'Okay, this is important to keep in mind. Our language is being shaped not just by people, but by the algorithm, especially on apps like TikTok. And I don't mean this in a hand-wringing, oh no, the kids are on the app kind of way, but in a very real, very specific way, because this tool and the incentives built into it are actively changing how we speak.
Adam Aleksic
00:02:31
I do think we should maybe be aware that language is an indicator of greater cultural changes and the way we're speaking tells us something about the way society is changing right now.
Audie Cornish
00:02:42
And that starts with the system at the center of all this. Here's how we understand the algorithm.
Adam Aleksic
00:02:48
Yeah, the algorithm is a personal recommendation system that distributes videos based on how their content is classified based on how users are classified and then they pair that information.
Audie Cornish
00:02:59
So what you're talking about with who wants to hear it.
Adam Aleksic
00:03:02
Right. But there's also an imagined thing intermediary, right? There is, as an influencer, I don't just talk directly to my audience as a broadcast journalist might talk to, you know, the perceived people watching the TV. I'm also talking to the algorithm, which is actually going to govern how my content is being distributed, and which is why I feel like I have to employ extra linguistic techniques or intonations or um so part of that is: This is the vehicle by which ideas and videos can even spread.
Audie Cornish
00:03:34
So let's talk about that in real practice. In radio, if we're going by medium, I wouldn't want dead air because somewhere somebody would flick a switch because they thought it was dead air. So I would make sure to always be talking no matter what. I also might talk a little bit slower because the perception is who knows who's listening. They could have all kinds of understanding of English. There's also a thing about accents. For the longest time in broadcast, people change their names, and also the idea was to have a very neutral, kind of regionless accent. And some of these things I have naturally, but some of the things I think obviously I was a little bit selected over time for having them. And it feels like the new medium is social media. And the new world of broadcasters are doing the same. There's ways that they've adapted that reflect, I guess, being rewarded in the new system.
Adam Aleksic
00:04:37
Yeah, some things are quite similar. You still need to kind of speak in a more homogenous bland accent so that more people can understand you. And I've interviewed a lot of creators who feel like they have to compress their speech into more stereotypically American sounding or stereotypically British sounding speech. And then even within that, maybe not consciously, but they're also selecting for their different influencer accents, whether they're an educational influencer or lifestyle influencer or what have you. Another thing that's different, though, is more commodification of attention the cultural logic of social media as we need we are actively grabbing your attention right now and this is different than long form video even it's short form video this vertical rectangle filling up your entire screen taking up the totality of your consciousness. You're trained to know that there's another dopamine hit right below my video if you want to keep scrolling so I need to make sure you don't even remember that dopamine hit I need make sure I have you at all times and that involves talking fast and stressing more words and using more uptalk and I'm putting something in the background so that you can be engaged by it all this kind of comes into play and it's a more deliberately active attempt to grab your attention.
Audie Cornish
00:05:45
OK, so for me, television, it's much more good morning, everyone. This is CNN this morning and today we're going to be talking. And here's where we start. And that's because Congress is and and sometimes I even joke around with my kids. They'll be like, oh, mommy, can you tell me something in the news? And I find I can code switch into it, right? Like, hello, everybody. Sammy is looking today for vanilla yogurt. Unfortunately, we're out and that's because mom forgot to go and buy some. And I didn't realize it until I was reading your book that that up and down and up and down is its own version of an accent. And what you're saying is online, there are many versions of that.
Adam Aleksic
00:06:27
Absolutely, you're good at these. I'm sure you'd be good at the online one too, but there's like—
Audie Cornish
00:06:29
No, I'm terrible. That's actually what this is all about. This entire interview is me trying to figure out why I can't crack social, because I don't know the accent. Okay, so just tell me some.
Adam Aleksic
00:06:38
There are different intonations you have to learn. Well, it depends on what audience you're talking to, right? The fact that I'm what I call an educational influencer. I'm stressing more words and a kind of a nerdy way to keep you watching my video. And I'm talking quickly. And there's like a social expectation of how I'm supposed to be talking. But then there's also like the lifestyle creator, like the hey, guys, welcome to CNN audio. And that's a completely different style. You're talking a different audience. It's more slowed down, but you're still filling up all the dead air. You're still adding these rising intonations, which make it feel like something is coming next. And you're still code switching into it. Like none of us actually talk like this in real life.
Audie Cornish
00:07:15
Can you talk about, is there a word that you had a lot of fun tracing its modern history? Like some phrase we use online or in real life, I guess, that came from the world online.
Adam Aleksic
00:07:31
Well, I mean, so much. I start the book with those examples of censorship avoidance speech. That's traditionally what's called algospeak. But the argument is—
Audie Cornish
00:07:38
So censorship avoidance speech, okay? Which I guess we all do to some extent.
Adam Aleksic
00:07:42
I mean, we've always been doing this and on alive is a euphemism for death and we've always been euphemizing death because we're all terrified of it. The word segs as a euphorism for sex is kind of like in 1948, the American author Norman Mailer tried to publish his book The Naked and the Dead and his publishers told him use the F word way too many times. So he replaces like all these instances of the f word with the word "fug" with a G and this to me is the exact same thing as seggs. It's like you have a medium which is preventing you from talking a certain way. You find a way to reroute around it and now we all really understand what's still going on, right? So —
Audie Cornish
00:08:13
I notice that with young women trying to talk about the issue of sexual assault, it'll be referred to as "SA" over and over again, and it took me a while to sort of figure it out. I just use the context clues like everyone else, but now I kind of understand what you're saying, what we're doing.
Adam Aleksic
00:08:29
Right, that's literally algo speak.
Audie Cornish
00:08:30
Yeah. Same thing with emojis, I guess.
Adam Aleksic
00:08:34
Totally. Yeah, there is like the eggplant instead of the penis. That's a classic one. But there's a lot of like stuff like this, the ninja emoji instead of the n word or people will find ways to sort of create minced oaths or circumvent something through illusion to a visual similarity or something like that. But yeah, that's that's only like kind of the tip of the iceberg of what I see as algo speak, right? I think it goes a lot deeper than that. I think algorithms are shaping every aspect of how we speak, from where words come from to how words spread to how those words get there. And, you know, you think about online words, the first things you're going to think about is like Rizz and skibbity and stuff like that. The classic.
Audie Cornish
00:09:10
So I'm thinking about how President Trump and the way he speaks, his vocabulary, is uniquely suited to this window of the internet, social media, how we interact. You've written about Trumpisms. Give me some Trumpisms
Adam Aleksic
00:09:27
Oh, sure, there's there's a lot that spread through the internet. There is make X, Y again, right? Thank you, X. Very cool saying sad as an interjection or saying many such cases or many people are saying this or putting believe me at the end of a sentence. And these are all like things that have somehow wormed their way into speech, first, ironically, as memes and then sometimes crystallizing into genuine speech. And I hear my friends unironically saying sad now.
Audie Cornish
00:09:53
Right, but the key is it has to be something that's not that sad, but would be personally detrimental to you. Like, I feel like there's an element of it that has to nod to Trump himself.
Adam Aleksic
00:10:02
There's an ironic context to how a lot of these things are employed.
Audie Cornish
00:10:05
Are you sure, or have we seen where the way we use it is no longer even related to Trump, himself?
Adam Aleksic
00:10:10
'Yes, so the context collapse is definitely real. There is new situations that these are applied to, and that's how memes actually stick and spread, that they can be applied to new situations. If it's a one-time thing, like, covfefe, that doesn't refer to anything new, it can't be readapted, it can't be reapplied. So there was a huge trend in Google searches to, for the word, covfefe, but then they immediately came crashing back down.
Audie Cornish
00:10:34
Right, so when Trump, accidentally tweeted this phrase.
Adam Aleksic
00:10:37
Oh yeah, 2017 or something.
Audie Cornish
00:10:39
Yeah, it was a joke for a week or two, but it's not slang now.
Adam Aleksic
00:10:44
The way things really enter our language is that they have applicability. They have they can be applied to new contexts. Definitionally, that is how we use language as a tool to apply to new context. So the stuff that sticks is going to be adaptable and it's going to be something that evades context. So—
Audie Cornish
00:11:03
Evades context. Whoa. Sorry. My mind is blown because my whole work life like everything is about like reminding people of the context and maybe it's another way that I'm not totally suited to to this phase of the Internet
Adam Aleksic
00:11:21
Well, the context is always changing and you're using it in correct contexts when you're using it. But, you know, in social media and on the TV and these are different contexts and languages used differently in these mediums.
Audie Cornish
00:11:33
Hey, etymology nerd said I'm using it right. Look at that. Look at us.
00:11:37
As long as we're talking and communicating, I think we're doing it right.
Audie Cornish
00:11:41
You know, the other thing is, I wonder if that's why the Gavin Newsom counter tweets where they are tweeting in the style of Trump catching the attention of everyone so intensely.
Adam Aleksic
00:11:54
'So I guess. One other thing to get to here is that I do think Trump is uniquely suited for the algorithmic medium or the way he talks was has always been to grab attention. He's trained in this from TV. He knows that about attention and its power and especially in the social media age, you just outcompete another person's idea through virality. And you are now more represented on social media feeds than the other person. You are now claiming a bigger presence, which is more opportunities for your supporters to donate. It's more opportunities to change people's minds. All of that and it doesn't... It doesn't matter if there's a counter message as long as you can out-compete the counter message so fake news can spread if there is no response mechanism.
Audie Cornish
00:12:36
Meaning the phrase fake news can be popularized because the algorithm kind of carries it along and the media doesn't have a counter phrase to that.
Adam Aleksic
00:12:44
Yeah, it's who's shouting louder and Trump's tweets literally feel like he's shouting and Gavin Newsom sort of tapped into that and there's is something to capitalizing words in certain ways or talking in unusual ways with it's been shown that Trump has a highly unique distinctive speech style in a way that differentiates him from previous presidents and this is actually very good for let's say generating comments, which pushes his speech further in the algorithm. And engagement is a metric that these platforms use to push videos and push tweets. So because Trump has a highly engaging manner of speaking, it actually is more suited for virality and that's why his tweets kind of look like that.
Audie Cornish
00:13:28
After the break, how chatbots speak is making its way into the modern lexicon. Or is it the other way around? We delve into it. Next.
Audie Cornish
00:13:41
'You have also started talking about AI itself and whether we are starting to adopt the language of AI. The reason why I wanted to talk about this is because AI basically is based on the mass theft of all the things we've said. You know what I mean? Like, the large language model, it's our language when we say the large-language model. And what is it about AI that makes it sound weird and kind of AI right now?
Adam Aleksic
00:14:09
Yeah, so AI is always going to have a flattened representation of reality, they can train on a bunch of stuff, but they won't understand language as it's actually used. And that means that there will be biased approximations. For example, we know that chat GPT says the word delve at rate way higher rates than usual.
Audie Cornish
00:14:24
We do? How do we know that? Like who's who's what etymologist was like, let's figure that out.
Adam Aleksic
00:14:29
There are some incredible researchers at FSU who have been working on this for a few years now. On a few angles, there have been researchers who have found that, yes, the word Delve in scientific abstracts has gone way up since ChatGPT came out. It's easy to say that, oh, just ChatGPT is using the word delve at higher rates than natural corpuses of English speech. But we now see that affecting our actual conversations.
Audie Cornish
00:14:54
Why isn't that a sign that people aren't just cheating and using AI to write things? You know what I mean? Like if that's the way the AI talks and you guys know that because you've been studying like how it's, what it spits out.
Adam Aleksic
00:15:07
They are. Yeah, they're definitely using it to cheat and but now that when you read that text, you're not sure whether a human right or an AI wrote it and there's an evaluation that happens in your head of is this English is this normal is different and you use words based on their availability. We've now evidence to studies have come out this summer showing that the spontaneous unscripted conversations we're having also have higher rates of the word delve. This is when there is no like possible AI use that could be occurring like live podcast conversations or recorded conversations. If I'm talking right now, maybe I'm using the word delve at higher rates, obviously I'm using it because I'm pointing it out but —
Audie Cornish
00:15:46
Yeah, and I'd be using it because hashtag public media, that's the kind of word we would use. But it's interesting because it's a word that conveys authority.
Adam Aleksic
00:15:55
Well, I think ChatGPT is trained to sound sycophantic and authoritative at the same time. So—.
Audie Cornish
00:16:01
Quite the combo.
00:16:03
Yeah, well, that's what it's doing. It's it sounds like it knows what it talking about, but it also is personalizing to you. It'll like mirror what you're saying as well. It's definitely not spitting out English as it naturally occurs. And nor does it actually have context and context again is such is it's the final thing. It's the most important thing We are contextual beings. We understand things in light of are their environments in their backgrounds But chatGPT does not and if you ask chatGP T to talk in African American English It'll sound very weird because it's not meant to it doesn't have the context and maybe it'll correctly replicate some things I oftentimes it will not um, so the fact that these these chatbots are speaking to us without context and yet, speaking at slightly different rates with the words at different frequencies that's going to affect our eventual uptake of language because we always just are ambiently shaped by what we perceive what we see is real. And if we think chatGPT is talking real English we eventually subconsciously replicate that.
Audie Cornish
00:17:07
Wait, so is AI talking like us or are we starting to talk like AI?
Adam Aleksic
00:17:11
I think we're in a circular feedback loop. It's a hermeneutic circle, where one, more human texts are incorporating AI, two, humans are reading both AI and the AI generated human texts, then humans actually start talking more like that, then we start writing real texts, real human generated text with more AI speech, and it's all in a constant spinning feedback loop. I think algorithms are doing something similar, I think it's the same conversation because algorithms are also based on machine learning. Both AI and algorithms are based on the same neural network kind of architecture where there's something goes in something crazy happens we don't know what it's called a black box and then there's an output and along the way there's a lot of opportunities for things to be represented differently than reality. The input could be biased.
Audie Cornish
00:18:00
Let me decode what you're saying, because for my mom, what I would say is what we're putting into it, the data we're putting into, it's taken in by businesses. They have their own priorities. So it's a business priority to encourage people to come up with like mermaid core and cottage core and like 100,000 fashion micro trends, because then you can sell and commodify those trends. So that means—
Adam Aleksic
00:18:27
Mermaid core is on the TikTok shop. Yeah.
Audie Cornish
00:18:29
Oh, it is? But what that means, I guess the thing that bothers me, because I don't want to sound like a Luddite. It's one thing for the medium to change how we do things, just like the radio changed things, and TV certainly changed politics. But I feel like everything about social media is about these businesses and what they want. You know, I joke sometimes that all social media is us doing free data entry for billionaires. And that's what bothers me.
Adam Aleksic
00:19:03
'Yeah, techno-feudalism is you are the product, you consume the product and you make the product. And that will basically, they just create the infrastructure for us to do their work for them and they gather data and that's how they're making their real money here. So this is something to be aware of and that the platform priorities are the ultimate thing that is overshadowing everything else and you can see how like Elon Musk tweets like tweaks grok so that it outputs what he wants more and there's a lot of evidence that he's doing that platforms are tweaking what they're recommending on your for you page based on what's going to make them more money at the end of the day.
Audie Cornish
00:19:42
The thing that makes me nervous is that on the one hand, I wanna opt out. I want to opt out. I wanna sound like myself. I wanna say things that I would say because I would say them. I wanna have conversations with my friends that are not like honestly a minefield of internet jargon. I also don't wanna be left behind by the culture. And become that person, you know, you don't want to be the last person saying Negro. That's what I usually use. Like there's certain language moves on and I'm both nervous it'll move on without me, but also really wanting to fight the power and be like, no, no I'm not going to be another thing for you to sell.
Adam Aleksic
00:20:31
That's such a valid concern, especially since language is, I think, occurring faster, that the change is being accelerated through the algorithmic incubator. I sleep with my phone in a different room. I think it's great, right? But I do think it is important to be somewhat aware because as the whole point of this book and what we're talking about right now is that the changes you can observe happening are actually indicators of where our culture is going, who has power in our culture, who's influencing our thoughts right now. And even if you're not on TikTok, if you were in a bar with your friends, the song that comes on is gonna be a TikTok song. The words your friends are using are gonna be TikTok words. So if you wanna understand why that song came on or why those words are being used, you have to conscientiously engage. But the key is to be conscientious with it. Detach a little bit when you're looking at social media. Consider, why am I getting this video that I'm getting? And you can still enjoy the video, but do it with a consciousness of what's occurring instead of falling victim to the algorithmic gaze where you just let things happen and don't process why they're happening.
Audie Cornish
00:21:32
Adam, is there anything that you wish people talked about or asked you that they haven't?
Adam Aleksic
00:21:39
I think we should be talking about vibes more. Ambient evaluations of reality. What we think is what we're feeling out. What we're feeling out.
Audie Cornish
00:21:45
Ambient evaluations of reality.
Adam Aleksic
00:21:48
'Yeah, like, okay, our vibe of whether or not it's okay to say the word delve, right? That's just like, that's a feeling you get. It's not like something you can actually articulate. It's it's pre-linguistic. But what you end up actually saying the linguistic output is based on the feeling you have before that. And I think it's very clear that our feelings are being shaped by the algorithm, and especially in a time when you don't know what two videos above your video was right you don't consciously start using words as much as you are surrounded by these feelings and politics is evolving through feelings first and facts later, right, the facts matter less than your ambient evaluations. Our memes, what's funny, this is like, funny is not a conscious decision you make to find something funny or not funny, it's based on your ambient feeling, it's on your vibe of what counts as funny, and this vibe is being shaped by the presence of the algorithm. We should be thinking about those vibes.
Audie Cornish
00:22:45
Another example would be the vibe session. For a while in the news media, we brought on economists to talk about the economy and how the economy actually wasn't so bad because this number and that data point and that point means things should actually be pretty good, but then the public had different vibes. And then you started to have news anchors kind of being like, it's the vibes session.
Adam Aleksic
00:23:09
Well, the economy is all just vibes.
Audie Cornish
00:23:09
Well, no, there's actual numbers there, right? Like you can think the economy is great because in your corner of the economy, it is great. And it can be terrible for another set of people. But somehow I feel like vibes can overtake.
Adam Aleksic
00:23:24
I think it's important to note that vibe session as a phrase was coined in 2022. That vibe coding is a phrase from this year describing how you don't go line by line to code anymore, but you ambiently evaluate what ChatGPT is going to spit out as code. That vibe check is from 2019, that vibe shift is from 2021, that we're using phrases like it's giving and mood and energy and all these words are going up and vibes is going up. We're always feeling things, right? Still—
Audie Cornish
00:23:49
I still don't know what vibes are. I know as I'm saying it, but now I'm like trying to define it and I don't how to say it. I guess it is definitely hard to define.
Adam Aleksic
00:23:55
I guess it is definitely hard to define yeah, it's it's the ambient experience of reality of what we think is real of how we engage what we what options we see as possible what we'd like what we find funny. So vibes are everything but it's that kind of environmental mood we get out of something and I think we're more aware of these feelings right now And this is reflected in our language We're more aware of these feelings because we're being affected more by ambient things rather than line by line logical things
Audie Cornish
00:24:25
Adam, thank you so much for speaking with us.
Adam Aleksic
00:24:28
Thank you for having me.
Audie Cornish
00:24:30
Adam Aleksic, author of "Algospeak: How Social Media is Transforming the Future of Language." That book is out now. I want to thank you so much for listening. Please review, subscribe, and share because it makes a difference, and we'll see you next week.