[ExI] Wikileaks.

[Anders Sandberg]

Eugen Leitl wrote:
>
> It's pretty well known that TLAs do FOAF network clustering,
> and of course if persecution level is dialed up sufficiently high
> such lists act as a self denial of service, which would be self
> defeating, unless you're into tabula rasa approaches a la
> Mr. Dzhugashvili. And these typically take some setup time,
> so there's plenty of warning.
>   

I wonder how sensitive these algorithms are to salting? Suppose each of 
us joins a randomly selected sinister group/mailinglist, or a random 
group in general, what would that do to the network clustering? From 
what I know of network clustering algorithms, this could mess up 
statistics with noise fairly well. Of course, TLAs and network 
sociologists are working on robust estimators and pattern finding. But 
even against robust algorithms, in some data mining domains it is known 
that an aware enemy can mess up classification (even when the algorithm 
is unknown).

It is fun to run this kind of network analysis. One can use it against 
one's enemies too - I am somewhat worried about "DIY illuminati 
software" making it so easy to mine and attack social networks that we 
get a lot of social noise.

> So apart from nutters, which are rare, I wouldn't worry too much

No, nutters are common. It is just that most of they do not matter. 
(proud recipient of two crazy missives today)

-- 
Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University