3. Reading Barabasi, A-L, and Bonabeau, E (2003), Scale-Free Networks. Scientific American, May 2003 Benkler (2006), chapter 7 Terranova (2004), chapter 2 http://www.barabasilab.com/index.php
4. Learning outcomes To understand the non-random characteristics of complex networks To apply theoretical models to the www and to social networks To consider implications for public relations
9. Barabasi and Albert (1999) The probability that any node on the network will be very highly connected to many others is VERY LOW The probability that a very large number of nodes will be connected very loosely or not at all is VERY HIGH Preferential attachment: new nodes prefer to attach to well-attached nodes
10. Huberman and Adamic (1999) Each website has an intrinsically different growth rate New sites are formed at an exponential rate PREFERENTIAL ATTACHMENT + GROWTH = ?
13. Implications The more popular you are, the more popular you become Niches are important Older nodes (sites) tend to be more popular than new ones, but only on average Money alone is not enough to guarantee future popularity or growth, but relevance and connection to already popular nodes can be
14. Clustering Sites cluster into densely-linked regions or communities of interest They link much more to each other than to nodes outside Clustering increases and intensifies as you move along the “long tail”
15. 28%: heavily interlinked: multiple redundant paths 22%: link to core, but not from core; new, or lower-interest sites 22%: link from core, but not to core; doc depositories or internal org sites 22%: cannot reach or be reached from core 10%: entirely isolated Benkler (2006): 248-9
17. Summary “ Bow tie” model repeats itself within clusters As clusters become smaller, attention is more evenly spread Very very few are receiving no attention at all