Supposing that the "Dead Internet Theory" is true or largely true, how can we act on that information?
post by SpectrumDT · 2025-01-27T16:47:01.338Z · LW · GW · No commentsThis is a question post.
Contents
Answers 7 Dagon 5 Dave Lindbergh 1 alexei@lebe.dev 0 jchan None No comments
The "Dead Internet Theory" (DIT) is defined by Wikipedia as follows:
The dead Internet theory is an online conspiracy theory that asserts, due to a coordinated and intentional effort, the Internet since 2016 or 2017 has consisted mainly of bot activity and automatically generated content manipulated by algorithmic curation to control the population and minimize organic human activity.
The DIT needs not be completely true in order for it to be a concern. Even if the DIT is just partially true, that is still bad. The worry is, of course, that these bots are not merely friendly-but-hapless C3POs but tools used by some powerful persons, corporations, or governments. I think it is obvious that if someone is willing to go to these lengths to deceive the general public, they probably do not have the interests of the general public at heart.
In the following, I will use the term "my DIT" to refer to the claim that:
In some specific non-trivial contexts, on average more than half of the participants in online debate who pose as distinct human beings are actually bots.
Let us suppose that we have one or more contexts for which we believe that my DIT is true - that certain areas of online debate are dominated by malicious bots. How can we act on this? What are the epistemological and pragmatic implications?
I am not able to fully validate all the information I read. If I am to form a picture of the world - and especially the complex "social world" of politics and economics, I must rely on other people who know more than I. How can I navigate such a landscape of untrustworthy sources? How can I be an effective altruist when I cannot trust my "senses" to tell me what the world looks like?
I would be grateful if you can point me to some existing articles on the topic, but original thoughts are also welcome.
Answers
I have strong doubts as to the effectiveness and secrecy of a single-headed conspiracy. My version of this is "Most popular, open, discussion or news sources are heavily biased and many participants (including reporters with a byline or chat hosts) have a POV or purpose different than the naive cooperative view."
Whether it's a bot, a sock-puppet (alt account), a thrall (idiot parroting some demagogue), or just a low-value uninformed "real" participant, 90% of everything is crap, and has been since before Ted Sturgeon identified the phenomenon in 1956.
It's not clear to me that this matters. The Internet has had a rather low signal-to-noise ratio since September 1993 (https://en.wikipedia.org/wiki/Eternal_September), simply because most people aren't terribly bright, and everyone is online.
It's only a tiny fraction of posters who have anything interesting to say.
Adding bots to the mix doesn't obviously make it significantly worse. If the bots are powered by sufficiently-smart AI, they might even make it better.
The challenge has always been to sort the signal from the noise - and still is.
↑ comment by meedstrom · 2025-01-28T16:59:05.511Z · LW(p) · GW(p)
I'm getting the sentiment "just sort the signal from the noise, same as always", and I disagree it's the same as always. Maybe if you already had some habits of epistemic hygiene such as default to null [LW · GW]:
The mental motion of “I didn’t really parse that paragraph, but sure, whatever, I’ll take the author’s word for it” is, in my introspective experience, absolutely identical to “I didn’t really parse that paragraph because it was bot-generated and didn’t make any sense so I couldn’t possibly have parsed it”, except that in the first case, I assume that the error lies with me rather than the text. This is not a safe assumption in a post-GPT2 world. Instead of “default to humility” (assume that when you don’t understand a passage, the passage is true and you’re just missing something) the ideal mental action in a world full of bots is “default to null” (if you don’t understand a passage, assume you’re in the same epistemic state as if you’d never read it at all.)
If you hadn't already cultivated such habits, it seems to me things have definitely changed since 1993. Amidst the noise is better-cloaked noise. Be that due to Dead Internet Theory or LLMs (not sure if the reason would matter). I understood OP's question as asking basically how do we sort signal from noise, given such cloaking?
I'll propose an overarching principle to either read things carefully enough for a gears-level [? · GW] understanding or not read it at all. And "default to null" is one practical side of that: it guards against one way you might accidentally store what you think is a gear, but isn't.
- For certain kinds of questions (e.g. "I need a new car; what should I get?"), it's better to ask a bunch of random people than to turn to the internet for advice.
- In order to be well-informed, you'll need to go out and meet people IRL who are connected (at least indirectly) to the thing you want information about.
In the following, I will use the term "my DIT" to refer to the claim that:
In some specific non-trivial contexts, on average more than half of the participants in online debate who pose as distinct human beings are actually bots.
I agree with this version, and I was surprised to see that the Wikipedia definition also includes the bit about it being a deliberate conspiracy, which seems like a strawman, since I have always understood the "Dead Internet Theory" to include only the first part. There's a lot of stuff on the internet that's very obviously AI-generated, and so it's not too far a stretch to suppose that there's also a lot of synthetic content that hides it better. But this can be explained far more simply than by some vast conspiracy - as SEO, marketing, and astroturfing campaigns.
If Dead Internet Theory is correct, when you see something online, the question you should ask yourself is not "Is this true?" but "Why am I seeing this?" This was always the case to some extent of any algorithmically-curated feed (where the algorithm is anything more complex than "show me all of the posts in reverse chronological order"), but is even more significant when the content itself is algorithmically generated.
If I'm searching online for information about e.g. what new car I should buy, there's a very strong incentive for all the algorithms involved (both the search engine itself, and the algorithm that spits out the list of recommended car models) to sell their recommendations to the highest bidder and churn out ex post facto justifications [LW · GW] explaining why their car is really the best. These algorithms are almost totally uncorrelated with the underlying fact about which car I'd actually want, so I know to consider it of very little value. On the other hand, I would argue, asking a bunch of random acquaintances for car recommendations is much more useful because, although they might not be experts, they were at least not specifically selected in order to deceive me. Even if I ask a friend and they say "Well, I haven't bought a new car in years, but I heard my coworker's cousin bought the XYZ and never stops complaining about it", then this is much more useful information than anything I could find online, because it's much less likely that my friend's coworker's cousin was specifically being paid to say that.
More broadly, on many questions of public concern there may be parties with a strong interest in using bots to create the impression of a broad consensus one way or another. This means that you have no choice but to go out into the real world and ask people, and hope ideally that they're not simply repeating what they read online, but have some non-AI-mediated connection to the thing.
No comments
Comments sorted by top scores.