Will the Science of Networks put paid to all the Snake Oil?

Attended the first two lectures at the Royal Society's Science of Networks conference today, from Albert Laslo-Barabasi and (Lord) Bob May. It was very refershing to hear Social Networking discussed with things like facts, rather than the more usual hype amd half-truths.

Barabasi in my opinion (literally) wrote the book on the subject, pulling together a lot of disparate strands in 2001, so it was interesting to hear what he had to say 10 years on about what was different or unexpected. My big takeaways were that:

(i) You have to differentiate between human and physical data networks. A lot of the initial theoretical work assumed humans would behave more like physical networks like the Web, and that there was less of a spread between richest and poorest nodes on the networks. (Physical networks have more randomness - but humans ones are far from random). It is becoming clear that Human networks are massively skewed towards a small group of very powerful nodes and a very long tail, that the differences (in followers etc) are huge, and that the "rich get richer" due to the preferential attachment principle (Most people tend to follow the most popular nodes)

(ii) This makes human networks more "scale free" than predicted, with far stronger power laws. They are driven by the dynamics of these power laws so reinforce early winners. They are also more "small world" than theory would have it (ie number of jumps from person A to B is low - "6 apart" for very large populations.

(iii) However, latecomers can make it big , but they have to have a higher "fitness" in the ecosystem. Think of the evolution social networks - Facebook is "fitter" than MySpace, which was "fitter" than Friendster. A later speaker showed how relatively small differences in user experience created huge changes in adoption, and that gives an indication of what "fitness" means.

(iv) Robustness is a real problem with scale free networks as the main nodes are so large. They degrade similarly to random networks if the main nodes are not taken out, but if they are the degradation is very much faster and worse.

I have ignored stuff that he covered 10 years ago, these are the big differences he highlighted. sadly there wasn't time for him to get into social networks proper.

Next up was (Lord) Bob May whosew ork I've also followed for 10 yers or so, his main area is stability in natural communities, but he used a lot of network maths to do it. He also applied these tools to the study of disease and to the study of biodiversity. His talk was on networks, ecosystems and a foray into how they can be applief to the financial system. takeaways:

(i) He pointed to a paper that seemed to use Game theory as well as network analysis for predator/prey network stability (Dunne and Gerwin 2008) which I haven't read (yet).

(ii) A interesting point about Sex Epidemic theory in Sub Sahran Africa - shouldn't happen on averag. However epidemiological network theory didn't take avccount of the small number of very active sex workers (shades of Barabasi and the Scale Free surprise) The key is then clearly to focus on these superspreaders and innoculate then, so gives a lower cost way to solve.

(ii) He made an interesting aside, to the effect that the worst thing you can have in these networks is to give the "superspreaders" carthe blanche to prey on the little people, and by the tone used (he has a marvellously dry wit) it was clear he was concerned that the major social networks today are going down these routes.

His next topic was one that has interested me this year, to teh extent that we have dine some work on it - namely using network theory a(and other tools like game theory) to understand Systemic Risk in banking. Big takeaway when they looked at teh banking network was that all the ingenuity that is going into ever more esoteric trading systems with higher and higher returns, but the overall banking ecosystem was being neglected as the big players got richer (shades of Scale Free big nodes and small prey above). The result:

(i) The banking networks is a huge networks of networks, and is massivley interconnected (and, I would conjecture, very scale free compared to other industries). So what happens if there is a shock? They built a model of 25 banks, and found that (as you'd expect from the above) if a big node went down the system collapsed - fast. This was especially true if the big banks had relatively small capital reserves compared to the rest of teh ecosystem - which is of course what they were doing in noughties.

(ii) He believes that the fact that all the banks are interlinked makes the whole thing literally "too big to fail", and he argued for splitting up the network so a failure in part A does not impact other parst - in essence a return to something like the Glass - Steagall act of the Great Depression that separated "casino" banks from high street banking.

(iii) Incidentally, the models showed that teh best thing for banks to do in the bad times is lend and not biuild up their capital, and to build up capital in good times (It's a Pharoah Story) - the very reverse of policy today. The models reckon that the current UK approach this is the best way to collapse an economy, by enrichng the banks and collapse the productive companies.

(iv) The other risk they can see is nanosecond trading is another emerging problem, as there is no governing of these systems. He made the allusion that if you put a half second delay in sending out emails it reduces virus spread.

He also made a good point re the economic "faith":

"There's an idea that some banks hold that there's an invisible hand which protects them, through market action, and not too much regulation, from bad outcomes," said May. "They're on very shaky theoretical ground there. The reason is it is invisible is thst it isn't there"

Actually its all too bloody visible, its called a bailout. The big thing that is missing in talking about banking "fitness ecosystems" is that in a Darwinian model they would all be extinct now, but have been allowed to live, unchanged. I was dying to ask him what he thought would happen next, given that - but they never picked me to ask the question

Sadly I then had to go back to work, but managed to return for the last talk, but that is for another post....... anyway, Sara Fletcher has done a blog of the day.

As I was about to hit "send" on this tonight, I caught an article from Malcolm Gladwell via Stowe Boyd: Gladwell notes that a new era of social activism on social networks is largely bollocks:

The world, we are told, is in the midst of a revolution. The new tools of social media have reinvented social activism. With Facebook and Twitter and the like, the traditional relationship between political authority and popular will has been upended, making it easier for the powerless to collaborate, coördinate, and give voice to their concerns. When ten thousand protesters took to the streets in Moldova in the spring of 2009 to protest against their country’s Communist government, the action was dubbed the Twitter Revolution, because of the means by which the demonstrators had been brought together. A few months after that, when student protests rocked Tehran, the State Department took the unusual step of asking Twitter to suspend scheduled maintenance of its Web site, because the Administration didn’t want such a critical organizing tool out of service at the height of the demonstrations. “Without Twitter the people of Iran would not have felt empowered and confident to stand up for freedom and democracy,”

......

The kind of activism associated with social media isn’t like this at all. The platforms of social media are built around weak ties. Twitter is a way of following (or being followed by) people you may never have met. Facebook is a tool for efficiently managing your acquaintances, for keeping up with the people you would not otherwise be able to stay in touch with. That’s why you can have a thousand “friends” on Facebook, as you never could in real life.

As readers of this blog will know, we have attacked "clicktivism" too, but from the Game Theory point of view of "weak tells", ie a person who supports a cause from a comfortable settee in London or whatever by turning their Twitter picture green is very unlikely to die for the cause. We get to the same place however.

Stowe's point is that the revolution is below this, at the "small - r" level.

I think Gladwell needs to look at the little r revolutions going on all around us, like the urban food movement, Grameen-style microloan systems, and others. This is where social tools will change the world, one weak tie at a time.

I think Barabasi and May would agree with Stowe's views, but also point out that if some nodes got too big (which being a human system they would probably predict they will without some form of governance), there would be major problems! But I would also love to hear May's acerbic wit on Clicktivism

As to the facts getting in the way of Social Media Snake Oil, I live in hope......