HFS: Human Flesh Search
Ian Alexander

Fig 1. Chinese root of ‘Human Flesh Search’

For a system to produce purposive results, it has to have been specified, designed, and tested against its requirements; it has to be commissioned and deployed prior to use; and it inevitably stays the same until it is deliberately changed, whether by redesign, “maintenance” software coding, or the updating of business rules if table-driven.

Only – all these laws are false, in the case of systems built on social networks and making use of the combined power of the Internet and human capabilities.

The International Council on Systems Engineering, INCOSE, has of course long defined a system as “a construct or collection of different elements that together produce results not obtainable by the elements alone. The elements, or parts, can include people, hardware, software, facilities, policies and documents…” (Eberhardt Rechtin, The Art of Systems Architecting, 2nd Ed., 2000).

It should be no surprise that the “parts” of an aircraft, or a bank, include both people and policies as well as hardware and software: these systems couldn’t work without rules to govern them and people to maintain them.

But these sober, Systems Engineering thoughts pale into insignificance beside the lurid accounts of Human Flesh Search that have appeared recently in the Western press. Western, for HFS stems from China – perhaps the first of many significant innovations from that dynamic economy and its many scientists and engineers.

HFS is an almost wilfully over-literal “translation” of the Chinese root (see Fig 1) which means something like People-Powered Search (you may recognise the first character as “Human” if like me you were ever introduced to Chinese script – and writing with a paintbrush – on a school Open Day). Or, as Fei-Yue Wang at the Chinese Academy of Sciences and colleagues in the August 2010 issue of IEEE Computer explains, HFS refers to “searches conducted with help from human users”, “often targeted at finding the identity of a human being”. This prosaic meaning didn’t stop the media from taking the shortest path to a juicy story. “Many erroneous notions have appeared on various blogs, on wiki sites and in media reporting”, comment Wang et al.

Among the dodgy definitions of Human Flesh Search published in 2008 when the story broke in the West, SearchEngineWatch came up with:

“finding and punishing people who publish material Web users consider inappropriate.”

The Murdoch site Times Online offered:

“digital witch hunts.”

And on a site which really should have known better, a guardian.co.uk blog entry in Comment Is Free described HFS as:

“an internet mob that hunts down real people online, then verbally abuses them and publishes the victim’s private information.”

Wang et al comment drily ‘Chinese-based sources offer broader definitions.” ChinaSupertrends.com defines HFS as:

“online crowds gathering via China’s bulletin board systems, chat rooms and instant messaging to collaborate on a common task. “

That task can be anything from fighting corruption, as in the famous “Outrageously-priced haircut” case, to uncovering fraudulent claims, as in the equally famous “South China Tiger” HFS case. And, yes, there have been some tragic cases, such as the lovelorn youth who claimed he was dying and needed to get in touch with his lost girlfriend – and then killed her when the HFS tracked her down for him. But hard cases make bad law.

The haircut in question came from a barber’s shop in Zhengzhou in Henan Province that charged up to 200 times the normal price, responding to complaints by asserting that it had “deep connections (presumably with the local government and the police)”. The local people responded effectively by using the Internet to identify who these deep connections were and to organize demonstrations in front of the barber’s shop: they also vandalized it, so there is some truth to some of the claims about HFS. However the case ended positively with official punishment of the barber’s shop management.

The South China Tiger case involved a hunter from Shaanxi province who claimed to have photographed the animal in question, which was believed extinct in the wild. The claim triggered an enormous amount of web activity; the photo appeared in Science magazine (“Rare-Tiger Photo Flap Makes Fur Fly in China”); and successive groups from the general public to professional photographers investigated and analysed the claim in minute detail. This led to the discovery of “the original calendar cover painting” from which the photograph had been created and edited (i.e. definitely fraudulently photoshopped).

The South China Tiger case is analysed by Wang and colleagues. They show that attention moved spontaneously from a single thread on the general site 163.com (with many entries and very many casual participants) to the professional photographers’ forum xitek.com, which cast serious doubt on the purported photograph. Attention then moved to tianya.cn, mop.com and sina.com.cn and half-a-dozen others which were less heavily involved. It took just over a month (from 12 October 2007 to 16 November 2007) from publicity being given to the photo to the discovery of the original calendar image.

The HFS systems in these and other cases were by no means all the same: they involved different groups of people, working with different types of search engine and different human “search” skills – humans are after all creative system “parts” and can fill in arbitrary gaps in machine capabilities, given sufficient time and motivation. HFS systems arise spontaneously, evolve dynamically in response to perceived problem needs – like finding someone or something, or applying photogrammetric skill, or searching official records, or applying pressure to local government.

The networks have unique properties, different for instance from blog sites or search engines. Both In- and Out-degree distributions follow power laws: “a few participants generated most of the citations, and a few participants got cited in these citations. We also found that out-degree dropped off much more rapidly than in-degree … fewer nodes that cited many other nodes than the nodes that received citations from a large number of other nodes.”

There were also (very) “few participants who appeared on more than one discussion forum. Although small in number, these nodes played a pivotal role in transferring and sharing information across discussion forums and often across HFS engines.”

In other words, an effective HFS system, tailored precisely to track down the South China Tiger, was formed by a few existing forums, several large groups of people with diverse skills (publicity and blogging; photography; the energy to search offline for tiger images…); and a few individual people who had suitable combinations of interests and hence could act as bridges between separate online systems.

With these “parts”, exactly the right system could form, evolve, set goals, acquire resources, and reach an accurate conclusion, rapidly and without either explicit financial investment or (ahem) requirements.

Share Button

Leave a Reply