Article Preview
TopIntroduction
Regular Web hyperlinks have limitations like unidirectional linkage, unverified destination, and issues related to relevance, reputation and trust (Leuf, 2006), some addressed by technologies like XLink (W3C, 2001). The stigmergic hyperlinks (stighs) we designed embody an alternative for some specific applications and propose some interesting paradigmatic changes.
“Stigmergy” means the mark of the work (Grassé, 1959), and it is a form of indirect communication, effective for some distributed control problems (Marco Dorigo, 2004).
Stighs are “stigmergic” because they communicate indirectly and, as collective, display emergent behaviors (that are not the result of centralized control mechanisms).
Stighs are also “hyperlinks” because they can look and feel like regular hypertext hyperlinks (Wardrip-Fruin, 2004). The main difference is that stighs have a “life” attribute that increases when they are used and decays at a natural pace, eventually down to a “death” level. This dynamic drives interesting emergent systemic behaviors.
We elaborate on the stigh’s architecture in the context of pervasive stigmergy, also in human-human relations and use other author’s taxonomy (Parunak, 2005) to frame our proposed object and look at it from a social epistemology perspective (Marsh & Onof, 2007), that includes Word Wide Web cases.
We discuss applications for stighs at two scale levels: the single Web page and the “generalized usage” scenario. For the Web page scale applications, we provide a demo at http:/stigh.org that can exhibit the behaviors that support decentralizing the finding of useful Web resources across a community of users and the automatic replacement of undesired destinations. These applications have similarities with recommendation systems (Linden, Smith, & York, 2003).
Regarding the generalized usage scenario, we discuss decentralizing search and stighs and the Deep Web. Decentralizing search would increase the “calculative capacity” (Callon & Muniesa, 2003) of Web authors and that would be a noticeable social change – this could be achieved if search engines could, at least partially, outsource some of their tasks, like crawling and/or parsing (Brin & Page, 1997), delegating on stighs or equivalent objects. The same delegation could be an approach to the Deep Web – the Web that the search engines ignore (Bergman, 2001) – because it is hard to generalize how to crawl it. Instead of generalizing, specialized stighs could handle particular cases.
The structure of this paper is as follows: first we explain the stigmergic hyperlink from a high level perspective, including the architecture of the individual agent and its environment, one possible taxonomical classification, and a social epistemology perspective of stigmergy and stighs. We then discuss possible applications from the single Web page scale to the generalized usage scenario. In order to better understand the internals of stigmergic hyperlinks, there is a detailed engineering section. After one illustrated example of stighs in action, we state some technological considerations and, finally, future work.