See also "Is the Tipping Point Toast ?" article by Clive Thompson in this month's issue of Fast Company. The long interview with Watts goes over some similar ground, describes how he tried reproducing the Milgram "six degrees of separation" experiment, and how he upset the (American) advertising industry. Watts is now on sabbatical at Yahoo working on his "Big Seed" marketing approach. Oh and you will hate the "order vs chaos" diagrams in the article.

I have given up reading books by journalists on scientific subjects unless highly recommended by someone I trust. The ones I particularly don't like are ones like a complexity book written by a journalist I read some time back that started something like "while the sunshine was bright on the desert landscape horizon I was just thinking of those lovely blue eyes from the last time I met scientist X in the gorgeous mountains while the snow ...." But Watts, a mathematician (he is Australian for those who like to know) is a true expert in this field and he is also very helpful. He has written to me a couple of times to help out with solving some of the implicit functions in one of his books (his "Small Worlds" book published by Princeton. He was very courteous and helpful. And he as also published an extremely well written popular book "Six Degrees" (at least 3 different publishers). What is particularly good about the book is that he is very careful to describe the problems associated with this type of mathematics and its limitations.
But your quotes here Dave certainly highlights big warnings for those undertaking social network analysis. I have once been seriously disliked for challenging such a diagram by repeatedly asking about the nature of the relationships, the implicit assumptions that have gone into defining and constructing them and how they were reading recommendations from them. Watts article is an excellent one to rebuff some of the social engineering practices of social scientist using these tools.

Cheryl, the "Is the Tipping Point Toast?" article is very interesting - What do you make of the conclusion? Mass marketing is still *likely* to be more effective on probabilities than targeted toward "influentials". (I have added the "likely" and "probabilities" because this is what Watt's work really addresses. I expect he doesn't imply that there are no circumstances in which aiming to a market will improve success). Another interesting point is that it would appear that exponential growth is extremely rare and can't be accounted for by standard models and certainly not available on by design. There is also an interesting paradox in some of the work claiming to be associated with eco-system landscapes.

I was involved on a project called the digital business eco-system (http://www.digital-ecosystem.org/) where a primary premise of the project was based on the power law distribution of scale free networks - in other words, that the core network nodes would act as influencers in the eco-system. Strangely, the project then took great pains to use social network models and other models relating to power law networks to determine policies to guarantee fairness in the eco-system and provide the "appropriate" behaviour and type of participants. I always thought this was more akin to creationism than evolution but it would appear from Watt's work it would be all in vein anyway.

Also Cheryl, I didn't see any "order vs chaos" diagrams in my version of the article and couldn't find any.

My favorite metaphor about retrospective attribution is the one of the nasty dolphins.
Common knowledge is that dolphins like humans and rescue reople in danger of drowning by bringing them back to the shore. Whereas: they just bring them in any direction, at random - but those who (just by chance) are brought to the cost survive and can spread the story of the dolphin who saved their life. And you will never hear the stories of the others ...
More seriously: we have this problem in statistical analysis of e.g. cancer studies, where patients die during the duration of the study. There are ways to handle it properly, but asking the survivers about why they survived is not enough.

Thats what i find so appealing about actor network theory, its the relationships that matter. Change theory just doesn't cut it for me, i know i behave differently depending on who i am with and what the innovation is about. Attributes don't stay rigidly fixed, context matters. I also like what John Law says of social science; when it tries to deal with things that are complex, diffuse, and messy, it tends to make a mess of it. Simple clear descriptions do not work well when the stuff itself is incoherent. The attempt to be clear, increases the mess.
I also found Talebs white swans annoying, an ok idea taken too far.
A good clear blog thanks, i enjoyed it.