Session 2: Search Engines and Power
Theo Röhle – Dissecting the Gatekeepers
Theo Röhle is a PhD candidate in media culture at Hamburg University. His dissertation seeks to establish Actor-Network-Theory (ANT) and Foucauldian concepts of power within search engine research.
Where does the power of search engines exist? One position of power is established in everyday discourse through images of anxiety and fear. Giving power a face however, tends to obscure the complex relations underlying it. As ANT suggests, there is no fixed source of power, just a temporary stabilization of a network.
Defining the actors in the power network, Röhle locates the search engine as intermediary between user and transparency, and between webmaster and attention. As Google enters the picture, it diverts all actions through its own network. From an ANT perspective, this makes it the obligatory passage point for both user and webmaster.
Instead of being a mere passage point, translation takes place. The PageRank algorithm establishes a shareholder democracy, where Google sees itself in charge of the interests of the Internet community. Furthermore, by providing a host of free webmaster tools and services, developers get associated and enrolled to index content that Google would otherwise not reach.
Through this enrollment webmasters enter Google’s disciplinary regime. Sites that do not comply with its webmaster code of conduct get punished by banishment, and webmasters are encouraged to report eachother.
A second regime Röhle identifies is governmental by nature, and based on the prediction of user behavior. In a manner of ‘risk management’ Google aims to gather intimate knowledge of people, in order to adapt its own behavior accordingly. A search engine, according to Röhle, is not primarily a tool to find information. It is an advertising system, translating user’s information needs into consumption needs.
One strategy to take would be to challenge the paradigm of technological closure. Links to privacy regulation are lost from the interface, while its minimal design falsely inspires a feeling of transparency. Furthermore, the right that commercial parties have on the use of our search data should be renegotiated.
Bernhard Rieder – Democratizing Search: Concepts and Challenges
Bernhard Rieder is assistant professor at the Département Hypermédia at the Université de Paris VIII. His research focuses on the theory, critique, and implementation of the concept of “delegation”.
The larger shift from information scarcity to information abundance has given filtering processes priority over distribution processes. This development emphasizes the increased importance of making judgments. Compared to earlier search technologies, web search is significantly different: It takes the human out of the domain of knowledge, in favor of the incessant, brainless crawlers.
The dominant ranking paradigm on the web being recursive link analysis (PageRank), guiding principles are popularity (the logic of the hit) and convenience. Commercial interests in the search market guide attention in the direction of already dominant sources and views on the web, leaving a large amount of sites invisible.
One of the strategies challenging this paradigm over the last few years is that of Wikia Search, a service that applies the Wikipedia principles of transparency, community, quality and privacy to search. Wikia Search is open source, it uses distributed crawling and relies on user feedback on search results. As the technology is still in beta, it remains unclear how this voting principle will unfold. Will the SEO and advertisement industry come in, and, in general, is there a ‘right way’ to present results?
Plurality, Rieder argues, is more important than transparency. Conflict, or lack of consensus need not stand in the way of living together between civil rights and normative decision-making.
The main challenges for democratizing search are the costs of infrastructure, quantifiable markers for ranking, and changing user habits versus software defaults. There is no easy solution, but elements of one can be found at the user, provider and interface level.
On the user side, searching skills and politics, the practice and politics of linking, and general informational ecology should be part of our education system. On the provider side, the focus should be on antitrust measures and the building of a public infrastructure based on crawling and linking.
On the interface level, Rieder notes several promising developments, such as natural language processing, advanced search initiatives, clustered search, and sliders to define the popularity-level of ranking. In general, we need better APIs and larger result sets, more insight in search habits and in the consequences of ranking, and a better conceptual grasp on search engines.
Rieder concludes with a call for getting people back in the loop, hybrid, plural and complex as they are.
Session 3: Making Things Visible
Richard Rogers – Deep Methods for the Info-Political Study of Search Engines
Richard Rogers is Chair in New Media & Digital Culture at the University of Amsterdam and director of the Govcom.org foundation.
After an introduction to the panel by Katja Mayer, Richard Rogers introduced the Amsterdam-based Digital Methods Initiative (DMI), a collaboration of Media Studies University of Amsterdam and Govcom.org Foundation. At DMI, researchers and programmers focus on developing research methods fitting to the technical specificities of new media.
When relating Digital Methods to the 1990s Virtual Methods research program, their different approaches originate from a different perspective on the relationship between the virtual and society. Whereas Virtual Methods focused on the embeddedness of the virtual in society, Digital Methods coined the term digital groundedness to research and demonstrate the way society is embedded within digital objects.
Rogers went on to compare and describe several methods to research ‘natively digital objects’ such as the link, the website, the engine and the sphere.
Looking at the research of links, social network analysis has been applied to the Internet fairly recently, studying the way websites or people are positioned within a network. Link studies by Digital Methods focus on Google’s view of links as votes. Through tools such as the IssueCrawler, a software tool that visualizes and localizes networks on the web, actors can be characterized and profiled based on their incoming and outgoing links.
Websites are commonly studied from a usability point of view or a SEO point of view, and site features can be judged, for instance, on their level of interactivity. An alternative approach would be to follow the medium and focus on the website as an archived object. Digital Methods uses the Internet Archive’s Wayback machine for this purpose, which privileges single site histories. The Google Politics of Tabs movie, online here, illustrates the methodology well.
It seems that googlization has made the engine an object for mass media critique and surveillance studies. The Digital Methods approach to study engines consists of capturing and saving search results, making it possible to look into ranking of a site or issue through time. The Issue Dramaturg was built to demonstrate the ‘drama’ behind information retrieval. Also, it uses a ‘Lipmannian device’ that queries the first hundred results for keywords on a certain issue – how far is it from the ‘top of the web’?
Spheres, finally, such as the blogosphere, often find there way into genre studies. Digital Methods sees spheres as demarcated areas, constructed by engines such as Google News (newssphere), or Technorati (tagosphere). It then becomes interesting to study and compare the prominence of certain issues within the various spheres.
Gon Zifroni & Tsila Hassine – Multipolar Search
Gon Zifroni is a designer and researcher at Metahaven Amsterdam, Tsila Hassine is an information artist and developer. Together they work on the Exodus search project.
Tsila Hassine kicked off by discussing several strategies to overcome the dominant ranking paradigm by web search engines, and the list as the dominant presentational metaphor.
Shmoogle.org could be Google’s best friend. Shmoogle allows you to perform a Google search, while losing the political and industrial weight of ranking. Google search results are presented randomly in a lengthy list without page breaks, advertising or hierarchy. Hassine notes that the design decisions made here are promoting the transparency of Google, and the access to web sites that hardly get found. The Shmoogle mechanism is one of pulling instead of pushing, ultimately leading to a broader web.
Another option to read information differently is to be creative with temporal organization. The IMAGE TRACER for instance, developed by Hassine herself and De Geuzen, archives Google image searches to track urls, appearance, disappearance and ranking through time. Saving images and metadata on a local server, a change of organization over time can be visualized.
Gon Zifroni then elaborated on networks and hierarchy. The scale-free character of the web promotes a ‘rich get richer’ mechanism, and network power is dispersed to the next connection. Linking within a community of consenting actors is an inward movement, which, on a network level, leads to disconnection.
Following this theory of ‘abundance of redundant relations’, Zifroni suggests a focus on the holes between high-density areas. These are more defining for their relation than links, and indicative of negative social capital. It is very hard to define a community on the web in a traditional understanding of borders, which is in turn promoted by search engine workings.
Zifroni proposes a theory about periphery, rather than the center. Attention should be called to the ‘bridging nodes’ in the network, the ones that are at the margin of clusters but might connect distinct spheres. With the Exodus research project, Hassine, Zifroni and others are looking at the periphery of information. As hubs and clusters exist on different scales, bridging across these scales will become highly important.
Discussion
The conference organizers had reserved the total of an hour for a plenary end discussion, turning out to be quite productive.
Aside from the mention of Wikia Search, the P2P alternative did not receive much attention in today’s talks but got elaborated on in the closing discussion. In terms of speed, the P2P engine will long remain in Google’s shadow. Tsila Hassine noted however, that regarding the quality of information we are looking for, it is far more interesting to think in terms of choice than in terms of speed. The idea of having several alternative models of search running in the background for specific searches is promising, considering developments such as Open Search and the Minerva distributed search project by the Max-Planck-Institut für Informatik.
Another subject not surfacing until the closing discussion was semantic search. In line with his earlier presentation, Paul Duguid argued that with search tools such as grep, we have moved into syntactic search, ignoring semantics. With information drawn from well informed sources, semantics would follow from the context. However, we move further and further away from that. Bernhard Rieder adds that not each resource is equally biased. Even in a list, we will find some way of consensus about value. Semantic search on the other hand, can give us something better than truth. Promising experiments are being done with experts classifying resources in folder structures, by way of a ‘soft ontology’.
An interesting remark made by an audience member, was that when we compare the industrial revolution to the information revolution, people saw their workforce institutionalized in the first. One of the speakers mentioned earlier that holding people together and engaging them in a struggle is very tough. According to Paul Duguid, the word user community gives a nice impression of a community united, but it is not, there is no collective user base. This fact also frustrates government intervention, as it is unclear on whose behalf it is acting. Furthermore, a significant difference between the US and Europe exists in a reversal of trust and mistrust put in institutions such as governments and corporations. The thought of a third party, a union or a body to lobby on a higher level could indeed be promising lest it, as Theo Röhle added, be less commercial and more political.
On the subject of organization, Richard Rogers noted to be glad not to have heard about the ‘wisdom of the crowds’ today. In contemplating what else to do with ‘everyone’, he expressed his faith in a shift from the panoptic to the synoptic model, in other words, to have many watchdogs watch the watchers.
With education coming up in most critical discussions on the future of search, the last subject of interest in the closing discussion was media literacy. It was suggested that, on the European level, media education is not a big issue. This thought was countered by Van Hoboken, noting that media literacy is in fact high on the European agenda. Indeed, it could be more prominently so and include more critical perspectives. Duguid mentioned that all sorts of literacy are taught in the American educational system, but the term media literacy is completely unclear. The notion changes with every decade; first off as ‘cultural literacy’, then ‘computer literacy’ and now ‘media literacy’. Education is very important, but literacy standards lead to shaky ground.
Publication
A publication following the conference is expected to appear in a couple of months.