Planet Telex

I often make the mistake of browsing the front page of Reddit. Humiliating, I know, but it's difficult to avoid when, as Ed Zitron would put it, the Internet is like, 5 apps. Where else can I even go to slurp up delicious popular sentiment?

One conspiracy theory that has become popular with the digital peasantry, particularly on Reddit, is the "dead Internet theory," which I'm sure you're no stranger to. In its simplest and least-objectionable form, the idea is this: a significant portion of activity on the Internet is actually produced by bots. This is almost self-evidently true. Everybody has experienced spam e-mails, robocalls, sketchy websites, and so on. Since the recent advancements in large language models (ChatGPT and so on) and image/video generation, this problem has predictably worsened. On top of the pre-existing bedrock of spam content, advanced machine learning has enabled the encroachment of realistic, fake content on all major social media platforms.

Like many good conspiracy theories, "dead Internet theory" is predicated on reasonable claims while also being ambiguous enough to easily expand in application far beyond the boundaries of its strictest definition. Notice how all of the important terms are really nebulous and difficult to define. What counts as "activity?" What is a "bot?" What does it even mean for online content to be "produced by a bot?" These questions are absolutely foundational to any of the claims usually associated with dead Internet theory, and yet they never really appear anywhere these claims are made.^[1]

Give the idea enough time to fester and we wind up with shit like this screenshot from a thread about a manufactured outrage surrounding Supergirl:

If you spend any time on Reddit you'll likely have seen claims very similar to these before. To be clear, all of this is just complete and total bullshit. I especially enjoy the snobby, highly-suggestive "Something like 50% of Reddit comments are bots." Wow, this person has just read so many scholarly, peer-reviewed research studies about this topic that they can't even remember the exact percentage.

I'll briefly cover why claims like this are untrue, but I won't spend too long since it's been covered well by others.^[1] Beyond that, I'll discuss the philosophical and social implications of the dead Internet conspiracy.

Why it's wrong

Starting with the statistical-sounding claims in the screenshot: as far as I'm aware, there has never been an actual research study claiming to prove that a certain percentage of comments on Reddit, X or Facebook are posted by "bots" pretending to be human. The comments in the screenshot above are likely misrememberings clickbait news articles like this, or this, or even this cnet article which boldly proclaims that "the Dead Internet has arrived." (TODO: This is actually a completely different series of news articles about a different study; this is different than the older news articles which were reporting on Microsoft studies)

Given the endless deluge of articles like these from mainstream news sources, I can't entirely blame people for drawing incorrect conclusions. Nevertheless, the claims in these articles are all completely different—both between each other as well as being completely different from the kinds of claims that are popular online (as in the screenshot). The underlying research findings from the news articles are always something like this: bot traffic has overtaken human traffic on the web (and somehow always "for the first time"). The reason for this is that network traffic is almost the only thing that can be reliably measured.

Think about it—how would these researchers "know" that the sampled online content is "bot content?" The form of the claim immediately implies some constraints on its possible origins:

• Where can a claim like this come from? Since ordinary users aren't presented with any way to distinguish between authentic and inauthentic content (say, a label that says THIS COMMENT WAS AI-GENERATED), the claim must come from some kind of insider who is able to "know" how the content was generated or uploaded—presumably, then, the tech companies themselves (although I don't think this is actually what people assume in most cases; more on that later)
• How do they know that content is bot-generated? Again, since the user interface doesn't show any distinction between real content and generated content, these hypothetical researchers must have some consistent means of distinguishing real content from fake content. Let's assume for now that, in the Reddit example, each post and comment comes with a property called isBot which can be either "true" or "false," and that researchers gathered statistics on the proportion of isBot = true compared to isBot = false posts over time
• Since the researchers are apparently able to distinguish between real and fake content so discretely, online content must be rigorously categorized, then. Let's assume that users posting content with bots are using a special program, and that when someone deploys a bot to surreptitiously post generated content, that the bot must go through a special bots-only door, and that when they post their content, that special isBot property is then always set to "true."

Take note of the major difference there: traffic, not content. Traffic simply means what it sounds like; a bot has visited a webpage, an app, etc. This is done for all kinds of reasons, from the mundane to the nefarious. One major thing that bots do, for example, is "scrape" the web, which just means that they visit webpages and gather up some kind of arbitrary data. A bad version of automated bot traffic would be vulnerability scanning, which is when bots go from website to website attempting to hack into back-end dashboards, user accounts, and so on. In neither of these extremely common forms of automated traffic does the bot interact with the platform's user base in a clandestine way (compromised accounts may indeed be used in a clandestine way, but more on that later).

Implications

What I really want is to explore some philosophical and social issues that I think are lurking underneath the conspiracy theory.

Experts, guardrails, and best practices: liberal metaphysics

• The expectation that everything falls into these neat "categories"
  • What is a "bot?"
    • If somebody uses ChatGPT to write a Reddit post for them, is that a bot? How do we intellectually contend with the huge amount of people who bought into the idea that "writing" with ChatGPT is "futuristic?"

The Internet is an epistemological disaster

• How mediated communication undermines social trust and democracy
• The vast majority of people are severely unprepared for the future

Behaving in ways that contradict your insane beliefs

Dead Internet theory as a coping mechanism

• Talk about Russia here
• Accusing everybody who disagrees with you of being a "bot"
  • This represents an imperfect, liberal counterpart of the totalized conservative simulacra
• How do people react when their belief in the conspiracy is contradicted by reality?
• Belief in the conspiracy prevents the subject from ever gaining crucial insight into how people "really are" or what's popular; it's a form of self-isolation
• By self-isolating, democracy is further degraded

How dead Internet theory flatters Silicon Valley

The increasing role of literacy in the social world

• A reasonable amount of media literacy is required for an individual to distinguish between real and generated content
  • Ability to recognize a variety of patterns is crucial for detecting generated text
  • We're already at the point where most people are apparently totally unable to distinguish between the Linked In-ish hype-speak of LLM's and real writing
  • Most people don't have the disposition to scrutinize the underlying semiotics of generated content: fast-paced corporate snake oil salesman talk; nauseatingly "cute" and objectifying video content of children, animals, etc.

How the open-endedness of the Internet contributes to this phenomenon (after all, it could've always been bots all along)

How extreme atomization makes solipsism appealing

What do we do in the future when dead Internet theory is truer than it is now?