Automation, algorithms, and self-censorship
I’m probably not going to turn this essay into a video and maybe not even a podcast. Not because of the theme or any of the concerns raised, but simply because I don’t know if I’ll have the time to do so this week.
A couple of my videos on YouTube got picked up by their recommendation algorithm.
Even by the most conservative “percentage of the audience who watched to the end” estimate, those videos had an audience that was at least one hundred times larger than my channel’s baseline. They’re still racking up views.
This raised a question I’ve been wrestling with since.
Not “how do I replicate this and make it happen again?”
Though that’d be an understandable thought.
Instead, I’ve been tackling the question of choice and self-censorship.
“Should I be trying to replicate this?”
Should I let the temptation of an algorithmically-driven mass audience on YouTube define and constrain what I do as a writer?
That’s obviously a loaded question with an equally obvious and loaded answer, but the journey between the two is what matters.
The creative process is made of the choices you make
Self-censorship is a vital part of the creative process.
The emphasis on “process” is intentional. In spite of what the name implies, acts of self-expression don’t occur in singular bursts. People keep using unfortunate terms like “seminal” to describe creative acts for a reason. They imagine that art comes in singular voluminous emissions followed by calm and well-being.
In conservative and religious cultures, that is then normally followed by shame and guilt, so perhaps it’s fitting that the rather obscene metaphor has become the standard, unthinking description everybody reaches for when they’re trying to describe a genuinely original creative or intellectual act.
To me that word has a smell. Whenever I see it in a text, that smell sits on the writing like a lukewarm blob. It’s a word that carries with it the act that precedes: there’s an implied dick pic in every paragraph that contains it.
As unfortunate as this popular metaphor is – and I can see why it’s popular – it’s also inaccurate.
I’m fine with obscenities in art and writing. It’s the fact that it’s so wrong that annoys me, and the coyness, obviously. There’s plenty of opportunity to write about genitalia when it comes to art and creativity. Resorting to a “teeheehee, oh my god, how rude!” Benny Hill metaphor is nothing short of cowardice.
It stands both as a shorthand dismissal of the weeding and winnowing, expression and self-censorship, that is the creative process and as an example of cowardly self-censorship that has stagnated into a petrified metaphor: the idae that creativity is the work of a singular moment.
This may well be an accepted truth, but it is not a fact. Creative works are the results of ongoing processes, not singular moments, and the processes are life-long – they are your life, interwoven with your life and reflecting it to the outside world. The singular inspiration coming out of the blue is a lie. A “eureka!” moment is usually the resolution or climax of an ongoing process that has stalled or reached an impasse, not the process in its entirety. It’s a climax preceded by lengthy work and succeeded by a no small amount of cleaning up.
Hang on. Sorry.
Maybe the metaphor isn’t that inaccurate after all.
It’s still an act of cowardice, though, an unthinking shortcut that lets you avoid thinking about what you say or what you intend to describe. Using the word is a choice, and choosing clichés like it without thinking means you’re letting an indistinct crowd of ghosts – the accumulated whispers of faded memories of texts read and forgotten – make your choices for you.
Tech is keen to automate those choices for you.
Navigating the automation and algorithms that have been pushed into every corner of our culture has become a core part of every modern creative process.
Even opting out of it – writing your thoughts out by hand is a choice made in the context of the laptop, keyboard, and software filled with various automations.
The choice to walk means something different in a world where the car is universal and ever-present.
Making poor choices is how you grow as a writer or artist
I’ve been struggling a bit with the choices I have to make in my writing.
Nothing too serious. My lack of familiarity with the algorithmically-compiled audiences of YouTube and TikTok mean I don’t really know how to address them. They feel like a mirage – a vision that’s ephemeral and changing.
From the perspective of a writer the algorithms abstract away the audience behind a veil of automation. The audience of a regular, non-algorithmic medium has a much more stable body that can be investigated with – essentially – a form of call and response.
For example, I do have a sense of my regular audience. Not in the analytical sense – I don’t actually know who you are – but I do have a sense of how my existing audience is likely to respond to my writing. I can’t predict what will play outside that immediate bubble or what will inspire a new reader to immediately subscribe or even email me, but I do have a vague sense for how the small community I’m in will respond when I try something in my newsletter.
I can speak to you, because you’re here.
My sense isn’t completely accurate, but I usually get the overall direction right.
I just quite often ignore that sense because I’m also writing this for myself – to help my put order to own thoughts. I know I can mostly get away with this because I’ve done it before. I know that the community I’m a part of has some tolerance for self-expression even when it doesn’t speak to them personally.
What I say to you is shaped by omissions. It’s not just a matter of choosing what to say, but since every decision involves picking one of alternatives, the omissions are always going to outnumber the choices. They are the outside that defines the frame of a painting.
Self-censorship is in this way a constant shadow of the entire creative process. Technology that affects or even governs that shadow changes the body casting the shadow – the work itself. It always has, because technology and culture are interwoven and integrated. A medium is always a technology of one kind or another.
But that technology used to be consistent, for the most part. It expanded opportunities, but it bounded that more open space with new sets of constraints. The printing press transformed text, but it also defined it through its limitations. This created a stable environment you, as an artist, could navigate and learn.
Algorithms have changed that.
Every call a writer makes to an algorithmic medium can have a different audience. Exploring the interests of your audience by changing your tone, style, or topic to see how they respond no longer works because the change algorithmically assembles a new crowd, some of which may be carried over from prior audiences, some of which may be new and strange and only interested in this hyper-specific new topic. Further exploration will lose them.
A particular kind of video will get you its corresponding algorithmically assembled audience. Make a different kind of video, and you risk getting thrown into a completely different kind of audience.
The self-censorship you do as a writer stops being about exploration and becomes a source of stagnation.
These algorithms favour repetition and routine. This is my fixed self, let the robots find it a place and auditorium. Nobody has chosen to be here.
The “algorithms” become audience automations. They transform art into a body severed in half.
Adding media is one thing, adding algorithms is another
I worried quite a bit about what to write for YouTube and TikTok. Much of this comes from my inexperience. What should I write about next? Do I dare try out different formats and structures? What approaches and angles can I take that might trigger a similar audience automation as before and let me preserve the creative body undivided?
All that worry did was leave me with a greater appreciation for newsletters, feed-driven websites, and podcasts in general.
I realised that the “algorithm” was also indirectly automating my self-censorship. Deliberate omission drives clarity. Fearful omission muddies.
Navigating the constraints imposed by a new medium is fun. Challenging and hard, but it is fun. If you enjoy writing, navigating new formal constraints to that writing means engaging with exactly the tasks and puzzle-solving that make the process fun.
Writing for voice means you need to reassess your sentence structure, pacing, and pauses. A pause in the right (or wrong) place is all it would take to turn the opening section of this essay on the word “seminal” into an exercise in camp.
Writing for something that will appear on the screen, even just a talking head, means you now have to be aware of the connotations the visuals might add to the text.
Facial expression while narrating the same line can be the difference between:
- “I’m making fun of myself.”
- “I’m making fun of you.”
- “I’m making fun of these people both you and I dislike.”
- And “I’m actually being quite serious now and not making fun of anybody.”
Thinking about the different ways a sentence can be interpreted is one of the tasks that makes writing enjoyable.
Following each of those thoughts is a choice and any choice you make is an omission – implied self-censorship – of the choices you don’t make.
Creativity is shaped through deliberate omission and seeing how the audience reacts, but in an algorithmic medium each choice risks creating a completely different audience.
This leaves you and me, as writers, with one of two, well, choices:
- You can ignore the algorithm and hope that the less algorithmic audience features, such as following, leaves you with a stable baseline that won’t be threatened by future alterations to the User Interface or notifications.
- You can stifle the urge to explore the form, stick with a set structure, format, and approach, and tailor your work to the audience automatically generated by the algorithm. Live by the algorithm; die by the algorithm.
Both are equally valid choices. Especially since not everybody has that urge to explore the form. Consistency can also be a virtue. Not everybody enjoys making choices, especially creative choices.
The audience, especially, should have the freedom to choose how much of a burden it takes on by viewing or reading.
Creative choices can be a burden
Having to consciously assess everything and constantly have to choose during every waking moment would be exhausting.
This is essentially why those of us with the neurotypes commonly labelled autistic more often than not find social interactions exhausting. The interactions, assessments, and responses that are automatic and unconscious to the neurotypical are conscious and take effort for many who are neurodivergent.
Unconscious processes can be a blessing – a relief.
It’s more than understandable that people, as audience, often don’t want to make choices.
Choosing media that is easy to unpack and reassemble in your mind, media that takes little conscious effort on your part, is a valid choice. I personally think that familiarity – choosing media you already know – is a better path towards a genuinely restful cultural experience than letting predigested slop ooze over you, but I understand why people opt for it.
It’s also a question of balance. Knowing that Josh Johnson has a consistent style and approach and that his YouTube channel has a predictable schedule is part of what makes his stand-up comedy a kind of cultural “comfort food”. You can usually guess in advance what the topic and structure of the video will be, but even though you know what he’s likely to joke about, you don’t know what the joke itself will be, and the framing stories that give his performances their heart is different every time.
Audiences respond to and appreciate a range of different opportunities for choice when experiencing media. Sometimes they like “difficult” media that needs active focus and engagement. Sometimes they want to lie like a slug under their weighted blanked and recharge while going “haha! He made fun of Elon Musk. So funny.”
The choices made by the creator are trickier to avoid.
Creativity made without decision or choice is slop
I wrestle with these questions and thoughts because writing without making choices – writing that’s assembled out of shortcuts, unconscious clichés, and predictable structures – is slop. The text and images generated by machine learning models may be the purest form of this kind of slop, but the genre has been with us since the first writer who recoiled from the intolerable burden of conscious choice retreated to the restful safety of an unconscious formula, uncritically adopted.
Making a choice can be painful. Poor choices stand out just as much as the good ones do. Many people don’t want to be noticed for being different. They have the impossible and paradoxical desire to be noticed for being the same as everybody else. They want recognition for their outstanding compliance.
They don’t want to be inventive on LinkedIn; they want to be noticed for playing by the rules of the platform extraordinarily well; they want to be uniquely not unique and be rewarded for exemplary blandness.
Slop is this automation of unconscious shortcuts, blended into a beige ooze, and poured into whatever shape is demanded by the algorithm or platform on that day.
It lets you be recognised for association with creativity while avoiding the risks of creativity.
Large Language Models are just one mechanism of many for accomplishing this. It’s a convenient automation, offered for a convenient price, but because slop is a larger phenomenon – not a unique consequence of a single technology – it’s a trap that we all need to avoid.
It’s easy to fall into repetition, to always play by the same structure and rules, to never vary your choices and always opt for ease and the sort of indistinct message that rewards lack of effort.
The creative act in modern media is a stormy, hard-to-navigate strait. On one side we have algorithmic media – the audience automations that abstract away the call-and-response exploration of your creative community – and on the other side we have algorithmic tools – the temptation to rid yourself of the pain of creative choice with the click of a button. Both jagged edges are shattering.
Thankfully, there’s more to this ocean than just those two cliffs.
The foundation of Josh Johnson’s work I mentioned earlier is his audience. Not his YouTube audience, though they are a big part of his platform today.
No, the foundation of his creative work are the audiences he has had to work with outside of YouTube. His stand-up is shaped by the people in the room where he’s performing. The Daily Show has a following built up over many years, and they define the work done on the show.
When he steps into those spaces and engages with those audiences, that exploration helps define and drive the work. YouTube’s algorithm adds to it and expands upon it, and I’m sure his team sees regular variations in the algorithmic response to the videos, but his work is not dependent on YouTube’s algorithmic traffic. Losing it would be a setback, but he would still be left with a foundation to build on.
This should make it obvious that search was yet another algorithm with the same risks and pitfalls as the platforms driven by recommendation.
Search was literally just a recommendation algorithm that responsed to key phrases. A review or blog post didn’t pop up in anybody’s recommendations, but through search its audience was nonetheless algorithmically tailored to that particular type of writing. Along with the dependence on algorithmic social media, most online writing was even more driven by and dependent on audience automation than your average YouTube channel.
Much of our media became dependent on the search and social media algorithms consistently delivering them a particular audience for a particular consistent and unchanging form of writing.
That went away because Large Language Models made it simpler to automate the slop than it was to automate the construction of an audience for manually created slop. Less lucrative, as it happens, but certainly simpler. Even without generative chatbots, none of this made for a stable foundation for our cultural and media industries.
Community is culture
These questions have an answer. The loose, informal communities that form around newsletters, blogs, and podcasts – the non-algorithmic media that remain – are the only audiences that have any chance of lasting. They’re the ones we, as writers, need to understand, and that understanding can drive our creative choices.
The algorithmically-assembled audience that comes from search and recommendations is an army of ghosts. If you’re lucky, they’ll come to you when you make something that resonates with them or with the moment. Otherwise, you’ll still have your actual audience – your actual community.
I’m not going to let YouTube, or the desire to succeed on YouTube, drive my choices. Newsletter comes first, then the podcast, then the YouTube channel.
That’s the priority of constituents.
That means that even though I’m planning to continue to deliver essays in text, audio, and video when I can, I won’t let the concerns of video limit what I write for the newsletter.
I’m not going to let the ephemeral algorithmic audience dictate what videos I make, and I’m sure the videos will find their own core audience in time.
This way I won’t drive myself into anxiety chasing the mirage of getting into people’s YouTube recs.