在手机上保存，获得更好体验
本字幕由TME AI技术生成
Welcome to the deep dive
Where the podcast
Where we take stacks of complex source material really analyze
What's going on and pull out the key insights
You need that's right
And today uh we are diving deep into something pretty fascinating
And maybe a bit unsettling
It's the we are language is changing like right now super fast
Yeah
And it's all being driven by technologispecispecially ally socisocial didialgorithms that let's space it kind of run our lives
These days absolutely the sources we've been digging into explore this thing called allogo speak
If you heard that too much
I've seen a flooding around
It's basically the language people cook up right sometimes on purpose sometimes maybe not just to deal with these online systems
Exactly systems designed to well control what get said or at least what get seen
So our misfind is around it or ways to play the game
So our mission today essentially is to unpack this whole phenomenon
It's someof like a new dialect or what linguis calall associalict right associaliis is just you know guguage is or so group
Yeah
But here the group is well potentially everyone online it's huge
It really is
And we need to figure out what happens
When this constant need to grab attention
And uh also to avoid getting censored or flagged starts actually changing the words we use every day how does that effect communication
Identity
All libit
And we've got a perfect almost canonical example to kick out of ah the word unalive
Okay Yeah
This one really jumped out at me from the research
It's kind of shocking
Actually we're not talking about some obscure internet lying here
No
That at all talking about middle schooars apparently they're using unalive in school essays like analyzing hamleor or just chatting with friend
And they see it is what a safeworld yeah like a classroom appropriate uopmism m for talking dedeath or suicide which is wild
It is well yeah
Know linguistically speaking finding ways to talk around death isn't new
Not at all right you point out dicease yeah
Disease es straight from land
Basically meaning departure yeah or even word die itself that actually replaced an older old english word swellton die came from old north
And it was likely a softer way to put it back
Then so people have always used ephhemisms for heavy topics like mortality that's ancient
But unalive feels different
What makes it stand out is just how fast it spread it's the speed
Yes
But critically it's what's driving the speed and the context
So the modern group unalive it seems to trace back to a uh twenty feteen s by their mman cartoon or me pretty meh okay so obscure origins zero
Then suddenly around twenty twenty one
Twenty twenty two
It explodes particularly on tiktok
And why people start using it specifically to get around the platforms automatic content moderation
The filters designed to catch talk about suicide
Exactly those systems are there for suicide prevention
But they often just block or remove content mentioning certain words
So unalive became away to well hacked the system lepanized ephmisism mmost
Yeah
You can see that
So it jumps from this niche mean stard to to use really widely
So showing ing
Off line line stschools oin less than ten years
Wow
Compared to dicease
Which took centuries to become standard
So yeah
Unalive really signals
We're in a new phase of how words are born in spread
It's etomology driven by these invisible algorithc forces
Consistently scanning
Sorting and sometimes rejecting what we say okay
So let's really unpact that mechanism this alguspeak you mention it started as this invasive coded language mainly to avoid censorship
And the source is say something like three quarters of americans have seen content flagged or moved
So it's common
This initial phase they call it the wacamall effect it's such a perfect description isn't it
But platform bans a word
People instantly find a substitumaybe a misspelling
Maybe a moji
Whatever the algorithm has to catch up
Ban the new term and boom another one pops up
Wachamall
We've seen this pattern before haven't way like way back
Oh Yeah
Think about leads speak from the old internet chat rooms in the eighties in nineties
Using numbers in symbols
Instead of letters like five one see one d three for suicide
It's the same emples just maybe definitely clluquque er
But the same basic idea finding ways to say things
The system tries to block
But today you're saying the game is different
The stakes are higher
It's not just about avoiding getting deleted right
It's evolved now it's often less about ouright removal
And more about avoiding suppression being shadow band uh the dretted shadow w and and actly
嗯
And this changes everything
It shifts the motivation from just being delevband let me pothis this without getting deleted exsomething
More 嗯
Promotional promotional
How so well the real goal for creators isn't just a post
It's to get scene to land on the recommended page
Before you feed the algorithmigolden ticket okay right maximum disticties so creators become paranoid
They worry they using certain worworeven
If they aren't explicitly ly hox might cause the algorithm to suddenly downgrade their content keep it off that main feed
They might get shadow band without any official notice
How do they even know if that happening they wash their analytics like hawmes if a video suddenly tanks
Or it's only getting views from people who already follow them now from the general preyou page
Yeah
They suspect the algorithm didn't like something
So they self sensor even more proactively
Precisely
The lack of clarity
The ambiguity of why something might be suppressed
Pushes creators to be overly cautious
Which means more uphhemisms more coated language more alggo speak
It's like walking on egg
Shows around a black box
You don't fully understand that's a great way to put it
And the funny thing is fugsshall this supposed ai sophistication
These systems can still make really basic mistakes
Which brings us to the the classic sunthora problem
H yes
The scun thora problem
Which next to maybe my favorite anecdote from the source material charles dickens muum um always brilliant so in twenty
Twenty
Twenty two
The actual chard dkens museum finds itself shadow ban on tiktok
Because
Because the name dicks contains a certain four letter substring
That the filter flagged as problematic no way
Seriously
Seriously
Their account dedicated to a funt intury novelist geit's caught by this blunt instrument filter
嗯
It's the exacsame issue that hit the poor folks in the english town of scun thorp years ago
When they couldn't sign up for email
Because their town name contained a flagged word
It's kind of mine buggling we're talking advanced machine learning ai
And it still tripped up by charles dicken's 's name
You think it learn context
You would
But new ones is hard from machines apparently lewhether it's a simple mple lender like that or complex moderation rules
People are clearly adapting their language constantly letaltalk about how they 're doing it
What does this creative lexicon look like okay
Yeah
The sources break down a few key strategies first up these sort of minststoths right ite sted oath
It's basically just tweeking aword
Slightly change a letter
Maybe swaa sound the meeting is still clear
But it slippps psed the automated filter we see examples like segs instead of sex
Often changing that hard ky or x sound to a softer g yep or things like lebyan using a dollar sign
Or even that really out there one curmit suwer slide as a stand in commcommsuicide that one is wow
It sounds almost absurd it does
And sometimes that slight silliness or obsurdity actually helps the term spread
Makes it more memorable
You know while still doing the job of the vasion ah and you mentioned earlier
This isn't fundamentally neor right people tweaking words to get pastigate keepers not at all
We taluked euphheisms for death
But think about censorship and publishing
Norman mayor back in in nineteen forty eight famously use a word fug instead of yeah well the other effort and henot will then naked in the dead just so we'd get published exactly same pressure
Same tactic avoid the sensor
Whether it's a human editor or an algorithm
The difference out was the speed mailers
Fug was a calculated move for a prinbook online
These things can emerge in spread globally in weeks even days okay
So minsed oths are one tool
What else the sources talked a lot about visual and acoustic tricks
Especispecially ua moish yeah
The emoji game is strong
The visual ones are pretty raraght forward
Right using an a moj that looks like the thing
You're talking about the egg plant for a penis
For the pedox
For um budcs or sometimes of vagina cherries for breasts
Oh
Those are pretty common knowledge
Now i think they become standarard neneshorhand
But then
Then you get into the more complex stuff
The acoustic evasiusing a mog es based on how word sound not how they look this is where gets really clever
Linguistically
It really is the examples given were things like the corn a moj being used for porn corn porn okay
I hear the rhmme were the grepe pooji for rape again a slat rime right
But then it gets even more abstract
The turtle of moje obused by some groups as a stand ding for the end word
Turtle n word
How does that connection even work
That's the point
The acoustic link there is much weaker may be basbason on turtle sounding vggly like ninja
Which itself has been used as a coded term
Which often intentionally oppiked outsiders
You have to be in the no like a secret code exactly the functions like sslang
Specifically like rmming slang
Think cocony riming slaang
Where you need the cultural context to decode it
Apples and pairs mean stairs
Zberry tarart means far okay
I see the parallel
The emoji
Itself turtle razberry means nothing to the algorithm or an outsider
But in group knows the sound of connection precisely and that vagness that reliance on sound or slent ryme
That only works in a specific context is what lets it sneak through the filters
While still communicating the intended meaning to the target audience super interesting
Okay
So we have minsed oths visual a moj ea ououstic
Momojs
好的好的
What about for really sensitive topics
Where it may be rhymes or or visual puns feel inappropriate yeah for things like sexual assaululto are violence strategy shift again we see a lot of initialisms and mattonomy so abbreviation right using sa for sexual saull is very common
Online it's direct
But refers the trtrigger word and mtonomy remindus that that and mattonomy is just using a related concept or symbol to stand in for the actual thing
So the source is mentioned people talk about getting red zoned
Red zoned
Yeah
It reers the concept to the red zone
Which is known on us college campuses
As that period at the start of the academic year
When students specififreshmen are statistically at highest risk of sexual assault
Ah okay
So using the name of the high risk period as a standding for the assault
Itself exactly alalws ws of the ten vivivors to discuss experiences and share warnings
Within their communities
Using language that resonates with them
But completely sidesteps the words that automated moderation systems looking for it creates a safer space for difficult conversations
So these are all really ingenious ways to deal with censorship
The wacamall game
But you hinted it earlier
That alguspec has become bigger than just that massively bigger
Yeah Yeah
The initial definition was definitely about evasion dodging the filters
Even now we really need a broader understanding of alguhospeak
It has to include any linguistic choice
People make because of algorithmake pressure
Even if it's not about avoiding a ban
Even then 是的
好的
Maybe especially then because the primary pressure now isn't just avoiding punishment it's seeking reward it's about maximizing attention
The attention economy that's the name of the game
So algust speak now and compasses how we talk
How we right
Even how we format things
Things all sigsigned keep the algorithm happhappy and crucially to keyou you
The listener
The viewer hooked to stoep you from scrolling away for even a second
And that pressure has actually created noticeable patterns
And how people speak online right like an influencer accent you could definitely call it
That there are some really distinct vercal features that seem optimized for attention
The most obvious one is probably up talk that rising pitch at the end of sentences
Like to he's a question
Exactly it makes statement sound unfinished
Yeah
Which psylogicically to compels you to keep listening
To hear the resolution tion
It's a freholding techchnie e basically keeps lilienener hanging on interesting
I thouthought it as a retention strategy
What what else others the gengenational difference in pause
People talk about the millennial pause uh
Yes
That split second delay before someone starts talking in a video
That tiny pause maybe just a second even less
But in the hyperfast world of platforms like tiktok that pause could be deadly for attention metrics
So jessie creators adapt any includly they tend to launch immediately into speech
No gap consistent stimulation for the very first frame
Any dead air
Any percceved desititch risks
The viewer swiping away
It's like a microbattle for attention in every single clip
Wow
The pressures intense beyond the rhythm and pauses what about the actual words chose
Oh
Absolutely there's a heavy reliance on highly emotive language lots of superative extreme phrasing
This is the best recipe
Ever you won't believe what happened my absolute favorite that kind of thing precisely everything is dialoped to eleven strong blowing life changing the worst mistake
Why because the algorithm
Favorors engagement and strong emotions by engagement bringer content the triggers and emotional response
Joy anger surprised discussed whatever is much more likely to get
Likes comcomments shares it's more likely go go vial al creators lelearn very quickly to the language that hits those emotional nerves
Hit the nerve get rewarded by the algorithm with more viability
So the language itself becomes optimize for emotional mintipulation
Yeah
You can argue that yeah or at least for emotional resionence
It's a direct feedback loop between language choice
An algorithmireward okay
So we've shifted from language
Avoiding censorship to language chasing attention
What does that due
How we see ourselves or identity
The sources talk about the explosion of microlabels
Yeah
This is fascinating the algorithc push for personalization for sorting users into ever more specific buckets
Has fuel this proliferation of hyperspecific identity labels
Especially around aestics s you mean like all the top core identitic
Exactly that core suffis is everywhere
Now it started way back with things like hardcore punk
Maybe norm core later on
But now it's exploded cottage core
Goblin core
Barbi core
Dark academia
Weird core
Trauma core
Angel core
There are on hundreds
Maybe thousands each one represents this tiny specific niche aesthetic and identity and the algorithm loves these nshes loves them
Because each label functions is incredibly precise medidata
It's like a perfect seo term for categoriing both content and users
So someone searches for cottacore outfits
The algorithm there's exactly what videtage or products to show them
It makes targeting incredibly efficient
Think about creators doing shin cocottacore halls
That label directly connects a specific ethetic interest with a specific consumer behavior for a specific retailor
It turns subcultures into searchable marketable demographics
Perfectly put one source
Even quoted someone from a platform
Saying subcultures are the new demographics
These labels offer a sense of belonging a way to define yourself
Yes
But they also make your identity incredibly easy to categorize target and ultimbly cell to okay that leads perfectly into the next section
This sort of paradox of spread algorithms are great at creating these nich filter bubbles
Using these labels right keeping people within their preferred content zounds
But the sources also say they're really good at pushing compelling content outwds
Even if that content or its language originates in a really specific may be even harmful nich
Yeah
That's the leakage problem
It's like the filter bubble is semi premiable
Like titinate groups develop their own ways of speaking their socialix or even fanalex like the specific vocabulary used by taylor wiwift fans is wifties okay
But if a piece of content or a particular term from that group is engaging enough funny shocking whatever the algernon might push it beyond the initial group 么
Eit enters the mainstream
Sometimes yeah
And this is where we can get tricky
Even dangerous
Look at the example of insel terminology okay in voluntary celebit communities known for ressaggeny right exactly highly problematic tic
Often extremist online groups
They develop their vocabulary
Plplaces
Fortchae or redit terms like maxing as it looks maxicc meaning
Trying to maximize your physical attracted right
I've seen that or being pilled like red pilled or black pilled
Referring to adopting certain harsh often analyistic world views
So terms from these really toxic french groups started leaking out
They moved from these close forms into more mainstream internet culture
Often picked up ironically here as part of mean culture you started seeing theteerical tick toks about being a walk pilled cardio max or something making fun of the obsessive self improvement focus guess
Yeah But
But re's the danger er se law remember that yeah
The idea that online without clear queues like tony voice
It's often possipossible to tell of someone is being genuineserious ous completely the irc ic still sarcasm doesn't always translate in text
So when these terms spread even as jokes
They can inadvtetely normalize ze underunderlying idea mean making fun of looks maxzine may still suddenly introduce ed spread the ore ore cepcept the obsesive focus on specific
Often obtataable maale beauty stand derive from an extremist ideology
The line blurr that's actually quite alarming the algorithm pushing a fringe language out potentially stripping it of context or normalizing harmful ideas through irony
It's a real concern
And it connects directly to another major issue highlighted in the sources
The incredibly rapid appropriation of language from marginalized communities
Particularly african
American
English or aae and queer specifically black ballroroom culture
Ok Yeah
Words like slt shade read souris giving
Exactly these terms historically have deep roots and black and quirer communities often serving as ways to build identity
Express creativity and even resist depression
They have specific meanings and cultural weight
But now they're just everywhere generic internet sying pretty much and the speed at which this happens now is unprecedented
So by algorithms that identify something as engaging or cool or funny
Often based on its origin in minority culture
So the algorithm spots the trend within a ae or bobroom culture amplilifiit it pushes it out to a much wider
Often white audience who then adopit frequently ripping it of it's its ggal al new on context and power
Sometimes even mispronouncing or completely misunderstanding it
The sources mentioned the word get
Oh Yeah
That's a prime example originating likely as an exaggerated pronunciation of goddam within certain black communities often used to express admiration particularly for a woman's figure okay
Then to get picked up goes viral and platforms like tiktok often used by non black creators
Sometimes in ways that feel like they're mocking aa speech patterns
And then worse people invent these completely false auomologies for it like those backryms grow you ate that or something exactly completely fabricated explanations that it raise the actual linguistic orgins and the cultural context
The speed of algorithm make difusion makes this preparation happen almost instantly
And the original communities lose control of their own linguistic creations
It becomes incredibly hard for them to reguuisor maintain the original meaning and significance of their own vocabulary
When it's blasted out to millions by an algorithm adopted and often work by people who have no connection to its roots
The words get divorced from their history in record time so okay
We've covered a lie
The vasion attention identity appropriation
If we pull all these threats together
What's the big picture
What's the main take away for you
The listener trying to make sense of this linguistic worirlwin
I think the core take away is that alga speak considered in this broader sense isn't just a quirky side effect of content moderation anymore
It's really a mirror refleking
How the entire attention economy is fundamently reshaping our language how how talk how words ds change how fast they chanined exactly it's accelerating linguistic evolution insteering it interrections dictaated by clicks
嗯
Views and algorithmic favorability faster and may be more powerfully than any communication technology i've ever had before
And there's that weird paradox you mentioned early right the algoriththm seem to be doing two things at once on one hand they push towards some origanization
Think about that generic influence or accent or how quickly trending slanin terms become ubiquous ous by everyeveryone everyone starstanding the same
Using the same viral phrases put on the other hand
They also drive this intense diversification
The explosion of all those hyperneage correspetics
The thriving fanalex the coded language within specific filter bubbles so simultaneously making us also unmore alike and splintering us into countless tiny highly specific tribes
It seems contradictory
But yeah
That's kind of what happening the algorithm sorts you feed you content
It seeed on that sorting
You adopt the language of that niche
Which further reinforces the sorting
It's a feedback loop
That shapes both collective trends and individual nshes and remember the gridom doesn't know it doesn't understand your authentic self
It just knows patterns
It predicts
Which you'll click on based on past behaviour in local egs
But when we ststart using language fed to us by that predict of sorting
Does that change who we are
That's the big lingerering tion isn't
It which leads this to that final provocative thought for you the listener to chew on we use language to build our identity to tell ourselves and the world who we are
But these algorithms are constantly nudging us towards adopting these hyperspecific labels
The courres the generational tags the fan terms
They offer sense of belonging sure community
But they also make our identities incredibly leggible
Predictable and ultimately marketable deep
So the question becomes how much genuine unoptimized individuality can really survive
When the very words we use to define ourselves are increasingly shaped
Sorted and served up for maximum algorithmaefficiency and less behonest commercial potential
What happens
When your identity becomes just another iata point and ily packaged online ld the languguyou use fitneicly into a prefine algorithmically approved box
That tension between finding connection through shared language online
And having that very language turn your identiity into a commodity
That feels like when the central conflicts of living in this algorithmggage
Wow Okay
That is definitely something think about about really powwerful
Maybe slightly cheilling place to leave it
This has been absolutely fascinating thanks for i'm packing all of that my pleasure
It's a complex and rapidly evolving landscape
For sure
It certainly is well
Thanks
Everyone fortining into the deep dive
Thanks for listening
We hope this gayou plplenty think think about jous next time for another deep dive into the sources
Shaping our world
我的
