The unintentional chilling effect that anti-AI hate has on marginalized creators
Attacking bot comments based on communication style doesn't hurt AI's feelings, but it does act as a red flag for humans who communicate differently
Welcome to The Late Diagnosed Diaries, musings of a late-discovered autistic author and speaker, former journalist, and eternally curious human, featuring educational explanations, critical explorations, and vulnerable realizations, all seasoned with a teeny little dash of feminist rage.
Lately, my Linkedin feed has been overrun with vitriol against the presence of “AI-generated comments.”
From screenshot-filled posts “exposing” the users suspected of leaving these comments, to nested replies attacking the commenters individually, it seems there’s a decent subset of the Linkedin user base that prides itself on being able to spot AI.
They look for certain tells, they say:
Repeating or echoing the wording in the post, sometimes verbatim
Being overly agreeable with the content of the post
Missing the underlying point of the post, or only replying to one part of it
Taking the words of the post literally, or not understanding an analogy
Stilted or repetitive sentence structure
Word choice that seems slightly “off,” or overly formal language
Don’t get me wrong: there are plenty of comments with these characteristics written by generative AI. Many are likely taking advantage of LinkedIn’s own “suggested” replies powered by AI, and others are trying to maximize engagement by using third-party tools to post automated comments.
But as I shared in a recent Linkedin post of my own, these bots won’t be deterred from commenting in the future just because you replied something sassy or shared a screenshot of their comment with a diatribe about how stupid they are.
But REAL people who see your responses might be.
Resorting to attacking, shaming, and ostracizing whenever we see linguistic patterns that don’t match our own will have little impact on the prevalence of AI or the people who use it.
But it can create a chilling effect on the expression of all kinds of marginalized people, including autistics, who communicate in ways that stray from that differ from the presumed default of the white, English-speaking, educated, neurotypical male.
And that includes autistics like me.
A Note On Marginalized Communicators
It’s important to acknowledge that it’s not JUST autistics who are impacted by this chilling effect though. There are plenty of other marginalized communicators who might be dissuaded from posting for fear that they will be targeted for language or linguistic patterns that stand out when measured against that “white, English-speaking, educated, neurotypical male” standard:
People speaking their non-native language, whose word choice may seem “off”
People using translation tools to post in a language they don’t speak at all
People new to an industry, who may not know the expected jargon
People of older generations, who may use different words, slang, or phrases
People from other regions, who may use different words, slang, or phrases
People from other cultures, who may use different words, slang, or phrases
People subject to abuse, who may deferential language to protect themselves
People using adaptive tools to communicate, whose fluency of expression may depend on the technology they use and how it integrates with other technology
Truthfully, almost any communicator who has had to adjust their language to navigate unequal power dynamics throughout their life is at risk of being falsely targeted by those trying to “expose” AI-generated comments and posts.
While I’ll be focusing on the risk to autistics below, keep in mind that the chilling effects extend far further as well.
Cringeworthy Criteria for “AI-generated” comments
As I reviewed the various “tells” I saw the self-appointed bot-catchers pointing to, I felt an uncomfortable pit growing in my stomach. Many of the characteristics they used as evidence, it seemed, could have been pulled from the diagnostic criteria that autistics are assessed with.
For example, it’s easy to see how autistics, known for our literal thinking, could be falsely flagged as AI for “taking the words of the post literally, or not understanding an analogy.
Literal thinking could also result in “missing the underlying point of the post, or only replying to one part of it,“ as could other common autistic traits like a predisposition to hyper-focus or having difficulty with abstraction.
But the two that hit hardest for me were the points about echoing the wording of the original post and being overly agreeable in response.
This is pretty much a textbook description of autistic masking.
What Is Autistic Masking?
Autistic Masking is essentially the act of trying to camouflage or blend in with those around you as a means of achieving acceptance, avoiding ridicule, and evading detection.
Masking is sometimes referred to as ‘camouflaging’, ‘social camouflaging’, ‘compensatory strategies’ and ‘passing’.
Autistic people have described masking as:
hyper-vigilance for and constant adaptation to the preferences and expectations (whether expressed, implied, or anticipated) of the people around you
tightly controlling and adjusting how you express yourself (including your needs, preferences, opinions, interests, personality, mannerisms and appearance) based on the real or anticipated reactions of others, both in the moment and over time
In some ways, masking can be seen as a beneficial tool or a strategy to be learned: something we turn on and consciously deploy when we want to encourage more positive outcomes and shift attention away from ourselves.
“One common fawning tactic among Autistics is mirroring: lightly mimicking the actions and emotions of another person, trying to meet the energy they are giving off so that they view us as normal and similar to themselves.”
— Dr.
Price, “Unmasking Autism.”
But it’s just as true to talk about masking as a trauma response, an unconscious means of self-preservation brought about by a lifetime of detecting that we’re being misunderstood, mocked, dismissed or ostracized when we show up as our authentic selves.
For most women and late-diagnosed people, in particular, we have been masking our true needs, interests, preferences, and more for most of our lives, both unconsciously and consciously. Even when we don’t know exactly what’s “wrong,” we can often sense that something isn’t working right.
And since the most ancient parts of our brains equate fitting in socially with surviving physically, those instincts take over: Do as they do. Say what they say. Acquiesce. Agree. Defer.
So, what does all this have to do with LinkedIn bots?
This whole situation seems like a deleted scene from Inception to me, so stay with me while I try to untangle the interwoven threads here:
When LinkedIn bros attack commenters they believe are using AI, they’re often seeking power. They want to appear smart and to be respected and celebrated for it.
When they do that, they’re using the same part of their ancient brain that drives masking: the part that tells us our social capital will determine our survival.
At the same time, their decision to “call out” others and go on the offensive that way also creates and reinforces the need for others to mask around them. Their power plays signal that they are not a safe person to be different around; they are merely waiting for you to trip up so they can turn you into a sacrificial lamb at the altar of their own social standing.
And that’s just the kind of person you need to be EXTRA unremarkable around. The kind of person that makes you want to mask or code switch even MORE. The kind you want to stay silent around. The kind you intentionally don’t engage with.
But that’s the thing about social networks: they need engagement to survive. Engagement is their currency. So when posts get minimal engagement, the social platform wants to fix it. So it uses tools like “AI-suggested replies” and AI-powered predictive text to try to lower the barrier to engagement.
Which results in comments that sound a little botty…
Which makes the bros go a little batty….
Which makes the maskers more masky….
And on.
And on.
And on.
Listen, I don’t know how to solve this whole thing, but it seemed like a thing worth pointing out.
Perhaps those who are hating on comments that sound a little bit different will think twice, knowing there are plenty of humans who sound a bit different to.
Perhaps they’ll think critically about why they dislike the difference to begin with.
And if nothing else, maybe they’ll realize that their attacks help create the silence that creates more of the bot content they dislike so much.
And the world could use a little more thinking and realizing, I think.