Content Is Not Content

There is something slightly ridiculous about writing a long essay about algorithms, attention, and the way platforms encourage us to make everything shorter and more immediately consumable.

Because this is, of course, a blog post. And apparently nobody has time for those any more.

Two-minute reads. Five things you need to know about the ten things. The thing you can understand before your coffee gets cold. So this one is already in trouble ...

It is too long.

Which is rather the point — and I’ll only mention that once more, at the end, once I’ve earned it.

I’ve been thinking about this through an unlikely combination of things: TikTok, sleepcasts, Jung, Žižek, artificial intelligence, and a seven-syllable philosophical word that apparently needs replacing with a computer game character.

Somewhere in there is an argument. Or perhaps a collection of things that have begun to look like the same argument from different directions.

Content

It is an extraordinary victory for marketing language that we now routinely use “content” to describe almost anything that appears on a screen.

Music is content.

A poem is content.

A carefully constructed piece of experimental sound is content.

A sleepcast is content.

A person talking into their phone while standing beside a car is content.

The word isn’t really describing what the thing is. It’s describing what the thing becomes inside a system. Content is stuff that can be distributed, measured, monetised, recommended — placed in front of a pair of eyes and converted into a number. The content isn’t necessarily what is contained in the thing. It’s what the system can extract from the thing.

Once everything becomes content, the thing itself starts to look strangely secondary.

The view that isn’t necessarily a view

I make a sleepcast. Someone watches it — except they don’t, really, because the whole point is that they eventually stop watching. Ideally, they fall asleep. From my perspective, this is wonderful. The sleepcast has done exactly what it was supposed to do.

From the platform’s perspective, something stranger has happened. The viewer has remained there. The video is still playing. Watch time is accumulating. A signal is being received.

I say: Wonderful, they’ve fallen asleep.

The platform says: Wonderful, they’re still here.

Both statements are true, and they mean completely different things.

The viewer’s attention has disappeared.

The platform’s measurement of attention has not.

So what exactly is a view?

A person consciously watching?

A measurable database event?

Evidence that the content “worked”?

Those aren’t the same thing. A view isn’t necessarily a viewing. A like isn’t necessarily liking. And perhaps the perfect viewer of a sleepcast is the viewer who isn’t there any more — the ideal consumer of the thing consumes it by ceasing to consume it.

A quite lovely little knot.

The mysterious disappearance of people

I’ve felt a smaller, dumber version of this myself.

I was posting regularly on TikTok, doing the things one is apparently supposed to do, and then, rather suddenly, the views dried up. Completely.

I stopped posting for a while. They came back. I have no idea why. But I have a theory. The internet has several hundred thousand theories. One is that consistent posting can start to resemble automated behaviour, and that a break somehow resets whatever suspicion has accumulated.

Maybe. Maybe not.

The important thing is that I can’t see the mechanism; only the outputs. And when the mechanism is opaque, humans do what humans always do: we build folk theories:

Post at this time, not that time.

This many hashtags, not that many.

Use this sound.

Don’t use that sound.

Post three times a day.

Don’t post three times a day.

We become amateur algorithmologists, inferring hidden machinery from its behaviour. Which is, it turns out, exactly what the machinery is doing to us.

Behaviour does not specify intention

Suppose the system sees a pattern and concludes that it looks automated. It doesn’t need to know what I intended. It only needs to recognise a shape. This is one of the awkward problems that appears in AI research around inverse reinforcement learning (IRL): the difficulty of inferring an agent’s underlying goals from its behaviour alone.

The problem is deliciously awkward.

The same behaviour can arise from different intentions, and different behaviours can arise from the same intention. Someone can give you a gift out of love, guilt, obligation, politeness, or a lost bet. The action looks identical. The internal reason doesn’t travel with it.

http://intention.It

This matters for the whole idea of alignment, which presupposes that there is something stable — a preference, a goal, an intention — to align with.

But the system never gets the intention. It gets traces. And then we do the same thing back to the machine.

We watch its outputs and construct a model of what it is “trying” to do. The platform infers me from my behaviour. I infer the platform from its behaviour. Neither of us has direct access to the other’s intention. We are both working with traces — making the whole exchange rather less like communication, and rather more like two sleepcasts, each guessing why the other has gone quiet.

The strange economics of scale

The system doesn’t have to hate you. It doesn’t have to suppress your work or decide you’re uninteresting. It doesn’t even have to notice you. There are simply too many of us.

At sufficient scale, individual collateral damage becomes statistically insignificant. A creator disappears. Another appears. No malicious intention is required.

An indifferent system is, in a way, more unsettling than a hostile one. An evil algorithm at least has an intention. This one just optimises whatever it was built to optimise, and doesn’t need to know you exist to do it.

And perhaps this is one of the strangest things about living inside systems like these. We keep looking for somebody.

Who decided this?

Who buried that?

Why did they suppress me?

Why did they promote them?

Sometimes there may well be people making deliberate decisions. But enormous systems don’t require an enormous somebody sitting behind a curtain. The machinery can produce an outcome without anyone intending that particular outcome. The absence of intention doesn’t make the effect imaginary. It just makes it harder to argue with.

Enantiodromia — or the Waluigi Effect

There’s an old Jungian idea that keeps surfacing here: enantiodromia, the tendency for something pushed towards an extreme to produce its apparent opposite.

The platform world is full of these reversals. Freedom of expression producing conformity. Infinite choice producing algorithmic narrowing. Entertainment curdling into labour. Escape from advertising becoming advertising everywhere.

The AI world has its own, rather more memorable term for a related phenomenon: the Waluigi Effect.

The appeal is obvious. It’s memorable. It’s funny. It’s easier to spell. And if part of the motivation for using it is that enantiodromia is a seven-syllable word that takes slightly longer to explain, we’ve stumbled into a perfect little example of the thing itself.

We optimise an idea for circulation, and in doing so, we change the conditions under which people are allowed to encounter it.

There’s a world of difference between: “This is a difficult idea. Let’s find a memorable way in.”

and: “This is a difficult word. Therefore people shouldn’t have to meet it.”

The first is pedagogy. The second is a kind of epistemic paternalism — and it has its own feedback loop.

We simplify because people supposedly can’t handle difficulty.

They encounter less of it.

It becomes unfamiliar.

Its unfamiliarity becomes “proof” that it was inaccessible all along.

So we simplify further.

A seven-syllable word is not actually much to ask of a person who can learn dinosaur names, medical terminology, football chants, song lyrics, usernames, passwords and the names of increasingly obscure Pokémon. It just takes a few seconds longer than Waluigi. Perhaps that’s exactly the problem the system is quietly solving for.

The difficult word doesn’t optimise. The meme does.

The algorithmic Babel

The same word can mean entirely different things to different participants in the system.

Take like.

To a viewer: I enjoyed this.

To a creator: Someone validated my work.

To the platform: A value of 1.

To an advertiser: A signal about an audience.

To a philosopher: A question about what liking even means.

Same button. Same event. Different worlds. An algorithmic Babel in which everyone uses the same vocabulary to refer to completely different things. This is where some of the strangest arguments come from.

The creator says: My work isn’t reaching people.

The platform says: Your content is underperforming.

The creator hears a judgement. The platform has only described a metric. A distribution failure gets misread as a verdict on the thing that was never distributed. Freedom to produce is not freedom to circulate. And circulation is power.

Interpassivity

Žižek’s idea of interpassivity fits neatly alongside this: systems performing activity on our behalf.

Television laughing for us. Autoplay deciding what comes next. Platforms remembering what we liked, what we watched, what we almost watched, and what people like us apparently ought to watch next.

We perform small rituals — like, subscribe, share — while an enormous amount of the actual work is outsourced upward.

The platform’s real trick may not be making us passive; it may be turning passivity into something that measures as participation. The machine can do the liking. The machine can do the remembering. The machine can do the recommending. And we can remain wonderfully busy pressing the buttons.

Three participants, three different events

The creator thinks: I made something.

The viewer thinks: I watched something.

The machine thinks: I measured something.

All three can be right, but still not be describing the same event. The creator experiences intention. The viewer experiences something. The machine receives a signal.

The trouble starts when we treat the signal as though it contains the other two. A number quietly stands in for a human experience, and because the number is visible, it borrows an authority that the experience underneath it may not deserve.

Ten thousand views feels like something. It is something. But it isn’t necessarily ten thousand acts of understanding. Or ten thousand acts of enjoyment. Or even ten thousand people watching. It is a number that stands in a particular relationship to a set of events. And the distinction is important.

Please watch this so you can stop watching it

Which brings me back to the sleepcast. A thing built specifically to help someone stop paying attention, uploaded to a platform whose entire economic logic runs on attention.

The viewer presses play. The voice continues. They stop listening. Stop looking. Maybe they fall asleep.

From my side, the work succeeded. From the platform’s side, a video accumulated watch time. Neither reading is wrong.

They’re just radically different descriptions of the same twenty minutes. The platform measures persistence. The human experiences disappearance. Possibly the smallest version of the entire problem, contained in one video file.

A small experiment in not knowing

At the scale of billions of interactions, the system can’t meaningfully understand any individual case. And it doesn’t need to. It only needs to classify, rank and recommend.

There will always be weird (unusuaL?) cases. People who don’t fit. Work the model can’t quite place. Intentions that don’t survive translation into signal.

That gap — between what I intended and what the system decided happened — isn’t necessarily a failure. In fact, the thing a system doesn’t know what to do with may be exactly the thing worth making. Not as an anti-algorithm stance.

I don’t think I could defeat an algorithm if I wanted to, and I’m not sure I’d want to. Algorithms are tools. The interesting question is simply what happens when we remember that the tool isn’t the thing it’s measuring.

A number is not a person. A recommendation is not a desire. A view is not a viewing. And a seven-syllable word is not an insurmountable barrier to human understanding. It just asks for a few more seconds than most things are willing to ask for any more.

Content is not content

There’s a particular irony in writing a long essay about the pressure to make everything shorter.

This one wanders. It repeats itself. It introduces a philosophical term that most people probably didn’t need. It spends several paragraphs worrying about the meaning of a button. It asks you to sit with things that could have been bullet points. It has failed rather spectacularly at being optimised content.

Which may be exactly why I like it. Because the argument isn’t that long is better. Or that algorithms are bad. Or that numbers don’t matter.

It’s that measurement is not meaning. A recommendation is not a desire. A signal is not an intention. And content is not necessarily what is contained in a thing.

Sometimes there is something left over. Something the measurement doesn’t capture. Something that doesn’t quite fit the category. Something the system can distribute without understanding. Perhaps that residue is where some of the interesting stuff lives.

Don't worry (😉), shorter pieces will follow this one ....

Each one will circulate, and each one will become, ironically, content.

But perhaps the longer piece can stay here underneath them. Neither because longer is automatically better. Nor because difficulty is automatically virtuous. But because sometimes an idea needs somewhere to be difficult. And perhaps the real question isn’t whether the algorithm understood what we meant. It’s whether we still know what we meant, once the algorithm has told us what happened.

That seems like reason enough to keep making things anyway.

#halFunusuaL #ContentIsntContent #Algorithms #AttentionEconomy #SocialMedia #TikTok #YouTube #DigitalCulture #AI #ArtificialIntelligence #Creativity #CreativePractice #Philosophy #Jung #Enantiodromia #WaluigiEffect #Interpassivity #Attention #AlgorithmicCulture #DigitalArt #ExperimentalArt #ExperimentalMusic #Process #Exploration