[ExI] Probability of being affected by terrorism

[Anders Sandberg]

Keith Henson wrote:
> On Thu, Jan 13, 2011 at 11:53 PM,   Anders Sandberg <anders at aleph.se> wrote:
>   
>> Actually, let's play around a bit with our assumptions and see what
>> happens. I think we have a pretty good model of terrorism being power
>> law distributed with exponent -2.5.
>>     
>
> http://physicsworld.com/cws/article/news/21465
>   

Exactly. I was even citing Clauset et al. in my blog. The reasons for 
this power law are more obscure. They have an interesting paper showing 
that terrorist groups also learn and have a development trajectory, 
http://arxiv.org/abs/0906.3287

Generally, any phenomena with a long/heavy tail has to look like a power 
law multiplied with a "slowly vaying function" due to mathematical 
constraints. There are also things reminiscent of the central limit 
theorem making them likely outcomes in a lot of systems.

The truly dangerous attacks are the ones that are obvious in retrospect. 
I suspect the next 911-like attack will not be based on any particularly 
exotic technology. However, the attacks to be worried about are likely 
biological - they have a much flatter distribution than the other kinds, 
and might spread arbitrariliy.

An interesting thing is that rationally for distributons with exponents 
between -1 and -2 we should spend nearly all our mitigation efforts on 
the far tail, ignoring the typical small attacks. That is of course not 
popular.

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