NetWorks

A good popular science book is hard to write. Linked by Albert-László Barabási exemplifies some of the common pitfalls. Its subtitle is "The New Science of Networks" and the front cover states its thesis as "How Everything is Connected to Everything Else and What it Means for Science, Business and Everyday Life".

A towering order, one which nobody short of a Feynman on a good day could deliver. Alas, although Linked is fast-paced and highly readable, its lack of technical justification for its biggest claims makes it ultimately unpersuasive. Does the published scientific literature cited in its footnotes prove the case? Impossible to tell. Barabási is clearly a smart researcher, as well as a good writer (setting aside distracting episodes of ethnocentrism --- did everything important originate in Hungary?). But his book would have been more illuminating if it had focused its light on a more modest set of goals, and if it had included more quantitative evidence and fewer hand-waving just-so stories.

Coincidentally, on 17 Jan 2002 Michiko Kakutani in the New York Times reviewed a different yet similar popularization, Predicting New Words: The Secrets of Their Success by Allan Metcalf. Like Linked, in Kakutani's judgment Predicting overreaches in its attempt to explain complex phenomena using a tiny set of principles. She finds its examples unconvincing and occasionally contradictory, and concludes "... the book as a whole is full of holes and flawed by a dogmatic and not always logical thesis."

Barabási's and others' research on networks has turned up some interesting, perhaps important, relationships, such as scale invariance and logarithmically "small worlds". But to leap from linkage statistics among web pages to grand postulates about physics, chemistry, biology, economics, society, and the universe? With only a handful of equations and even fewer graphs of real data? And with no discussion of the limitations and weaknesses of key hypotheses?

I'm skeptical --- as I try to be with all freshly minted theories-of-everything.

(see also VulnerableTheories (17 May 1999), ScienceVersusStampCollecting (20 Jun 2000), ScienceAndPseudoscience (6 Oct 2001), ErdosNumberz (13 Jul 2002), ...)


TopicLiterature - TopicScience - Datetag20030120



(correlates: NaturalPhilosophy, ThreeThoughts, AnnotationPunctuation, ...)