A couple of years ago, a journalist friend asked me whether I knew of any political scientist working on the political consequences of chain email. I had to tell him that I knew of no-one (the Americanist colleagues I asked didn’t know of anyone either). It looks as though I was looking in the wrong places. Computer scientists have “been on this for a while”:http://www.pnas.org/content/105/12/4633.full. In the linked article, David Liben-Knowell and Jon Kleinberg try to trace the processes through which chain emails are propagated.
bq. To reconstruct instances in which specific pieces of information spread through large, globally distributed populations, we analyzed the dissemination of petitions that circulated widely in chain-letter form on the Internet over the past several years. The petitions instruct each recipient to append his or her name to a copy of the letter and then forward it to friends. Each copy will thus contain a list of people, representing a particular sequence of forwardings of the message; and hence different copies will contain different but overlapping lists of people, reflecting the paths they followed to their respective current recipients. … The main chain letter that we analyze is based on a widely circulated petition from 2002–2003 claiming to organize opposition to the impending war in Iraq. We obtained copies via Internet searches of mailing-list archives in which they were publicly posted; these searches resulted in 637 copies with distinct chains of recipients, representing nearly 20,000 distinct signatories in aggregate.
bq. Although both petitions in fact had their origins in hoaxes and naive misunderstandings, as a large fraction of the most widespread Internet chain letters do (22, 23), this fact is immaterial to our purposes … Inspection of the few messages that contained intact addressee lists indicates that recipients generally forwarded copies of the letter to a large number of other individuals. … IThe tree reconstructed from the data, however, reveals a structure that is very different from the picture suggested by simple epidemic models: the median distance to the root over all nodes is nearly 300, and >90% of the nodes have exactly one child. … The tree reconstructed from the data, however, reveals a structure that is very different from the picture suggested by simple epidemic models: the median distance to the root over all nodes is nearly 300, and >90% of the nodes have exactly one child.
I’m less confident than the authors that these emails are typical – it seems plausible that hoax emails will have a higher discard rate than regular ones, and perhaps differ significantly in other aspects of their dissemination. But it’s still impressive that this work was done at all (and a signal that political scientists who wish to exploit data from the Internet are going to have to tool up with a whole new set of methodologies if they wish to take advantage of the possibilities that these data offer).