"Don't mention the X!": the XKCD effect, naughty pictures, and information hazards - accidental "thinking loud" thread. TL;DR: what are the conditions for harmful ideas that act just through mention?
I came across https://xkcd.com/696/  and (as was noted already in 2010) it caused there to be Google hits for the nonexistent cases. I noticed that now there are even actual (if tongue in cheek) examples of all of them except strip podracing. Which is as it should be, I think.
This is a relative to the Streisand effect (let's call it the XKCD effect): calling attention to the nonexistence of something informational causes it to emerge.
Note that in some cases an actual instance of a previously nonexistent entity is created, either by (1) the act of mention or by others mentioning/discussing the mention, or (2) by someone creating a referent.
"Nobody talks about X" is of course the classical case of #1 (colorless green ideas, anyone?)

(Although usually that means "I think we should talk about X more!", sometimes spiced up by hinting that They don't want you to talk about X, so it must be cool/important/subversive).
The classic example of #2 referent creation is of course "Internet Rule 35", which made me make a naughty picture of dodecahedra: https://flic.kr/p/7riNkp  (and led to some debate about how many referents one can create implicitly)
From an information hazards standpoint this seems fearful: speak of the Devil and he shall appear. But this only works for harmful info that is *easy to make but hard to independently discover/reinvent*. Usually people handwave towards "attention hazards".
Empirically, there have been cases where showing interest in X triggers adversaries to investigate X, leading to bad outcomes. But here X has usually been technologies, or facts that were overlooked. I can't recall any good example of a hazardous #1 XKCD effect.
The closest would be a pure idea like computer viruses (but Cohen 1984 coined the term and made it visible, yet there were existing examples going back to the early 70s, plot point in several 70s sf stories, and it is described in von Neumann 1949).
Another example is "now you can't unsee it" where recognizing a pattern causes harm (or, usually, disgust/amusement). But this is dependent on an external piece of information rather than being invented whole cloth.
Are there any good examples of single-source ideas whose *mention* itself actually have produced an information hazard?

Seems *pure* mention does not work, there must be a semantic link to other things (those colorless green ideas strike again in their furious sleep).
The reason I am pondering these things is that fear of attention hazards seems to be profoundly corrosive for epistemic systems: bounding their power and detectability makes it much easier to find countermeasures without tearing up important institutional tools for thinking.
When you do not allow yourself to think or communicate about X, is that because you did a proper risk assessment, or just let a jumble of intuitions paint a scenario? And if you think assessment causes risk, is that because you have good reasons to think it, or the intuitions?
You can follow @anderssandberg.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: