Sunday, November 15, 2015

We need a scientific, epidemiological, study of ISIL

Simple infection model (F2 from Rasu 2014)
The terrible events in Paris really bring into focus the need to consider this ISIL phenomenon from a scientific and epidemiological point of view. ISIL is spread, to a significant extent, through social media and the internet is almost certainly used to coordinate attacks in the West.  The large Internet companies make their money by detailed algorithmic analysis of the information they have - and Google scans all the emails it its system. It's therefore high time that significant resources were devoted to the scientific study of the epidemiology involved and how to disrupt it.

As an start we can consider a six state path:
  1. Innocent
  2. Curious
  3. Sympathetic
  4. Radicalised
  5. Engaged
  6. Terrorist
  7. eXited (either dead or inactive).
We can write pij to be the probability per unit time of moving from state i to state j. For simplicity we might assume for the moment that only transitions between adjacent states are allowed.

Clearly if the only transitions are to higher states than unless p67 is very large indeed (terrorists give up or are killed very quickly) you will get more and more terrorists. Making p67 high is the focus of military operations, but if at any point in the chain we can get the flow going backwards and keep it there then the pool of Terrorists will decay exponentially. Even if we just have p43 greater than p34 then eventually (with a fixed population) the supply will decline. That BTW is why it is so important for ISIL to get women to come to their territory, because if they keep having babies then in 15-20 years ISIL will not be dependent on external recruits and be much harder to beat.

There is so much known about persuasion techniques on social media and otherwise that it should be very feasible to target people especially at stages 2 and 3 (C and S) and work hard to increase p32 and p21. At that stage people are, I think, using quite visible social networks (like Facebook) and it's pretty easy to see who they are, what they are interested in, and who they are following. Almost everyone has someone they will listen to, and a systematic, scientific approach could be made to figure out what are the most effective messages to send to each individual.

The Alan Turing Institute, whose launch I attended on Wednesday, would be an excellent place for such research. And the internet companies should fund it.

The ATI launch was an excellent event BTW. Andrew Blake, who I've known since we were undergraduates at Trinity, is truly world class and under his leadership the Institute will be. I also had the pleasure of meeting one of Alan Turing's nieces and one of her daughters (who read maths and is now a primary school teacher).

PS Interesting article in Knowledge at Wharton relates to some of this: talking about a coordinated soft-power campaign against ISIL.

No comments: