With: Siva Vaidhyanathan, Martin Feuz  and Esther Weltevrede

Moderated by Andrew Keen.

Moderator: Why does no one talk about money?

Vaidhyanathan: Google only loses money. They have an interest to keeping people interacting with the Web. As long as you are interacting with the web, they can track you via cookies and that puts more data in their database. It is a clear but third degree connection for creating revenue. It also has interest in data- and text accumulation. It hopes to create a real text-based search. In terms of Google search; global and local are not really put to for example; Google books. This already biases the search results.

Weltevrede: It also depends on your perspective on Google. For me it is interesting to see how it works. How does it organize and present the information we get.

Vaidhyanathan: nobody is going to Google for the ads.

Audience (at Weltevrede): you were depending on the Google translation results?  Isn’t that tricky?

Weltevrede: indeed,  Google Translate is still in beta version. However, human rights is such an important term that one can assume that it is translated well.

Audience: how about methods? It is difficult to pose yourself against the machine. All of us here agree that searching sucks and that Google is bad and commercial. So I’d like to have some reflection on methods in order to be critical against searching and how they relate to your research?

Vaidhyanathan: Google is hard to study in traditional way. I do my best to keep to fuzzy, flabby arguments of narrative and argument. Opacity is the problem of Google. You cannot research is without using it. You risk becoming a banned user. You have to warn Google about your research, in which you may alter the results.

Weltevrede. I agree, I want to add that you can study the inner workings by looking at output, you can tell a lot about that

Feuz: it is an attempt to look at temporal relations: You have to try and fund ways to be able to ask these questions.

Moderator; What I do not understand is the connection between the most opaque company ever which are still fetishing transparency.

Vaidhyanathan: it does not fetishize it; it leverages it. We do the work for Google, we provide the information and content Marx would scream at this notion. We are all very happy to do it (user-generated content). It is a better environment than we used to. However, we have to grasp the workings. Maybe we are very content with our relation to Google.

Weltevrede: it is also what building tools you can get out of Google. You can make use of the giant – building on Google; let Google work for us again.

Manovich (audience): I have difficulty to see your (Feuz’s and Weltevrede’s) results as research. What is the sample size? Where is the statistical data? You haven’t looked at the interdependencies of the variables? So what kind of science is this? If these things are not clear, these results are not meaningful.

Feuz: there is a difference between types of research. In the kind of research I did, I worked 4 month in a team gathering data. The amount of data we needed was already overwhelmingly large. You have to keep in mind that the thing is really temporal.

Vaidhyanathan (at Manovich): Is it not very expensive what you do? How can you do this?

Manovich: Most things are done in open source software and only takes five minutes.

Rogers (audience): Responds to the question by Manovich on what kind of science this is: it is diagnostics! Are local Googles furnishing local sources? It is a kind of critical diagnostics to see how Google works, and to see at the implications.

Manovich: Is it then issue exploration to be followed by hypothesis?

Moderator: I live in Silicon Valley, There is more skepticism there about Google. They cannot fight the real-time twitter economy. What is the relevancy of Google right now? What are your thoughts about this? Will it remains strong?

Vaidhyanathan: I am very bad at predicting. For the sake of my book, I hope they stay relevant? The rapid changes of Google have made me realize I must not write about current companies anymore. You have to keep in mind, though, that the real time web is not that connected (yet). So much of what Google tries to do is to satisfy the Cosmo-elite because this group makes the choices and the critics. What are the initiatives that Google has in India, China and Brazil? That is a more relevant development to look into.

Feuz; we researchers cannot cope with the patterns of change – they can adopt fast, so they will survive.

Lecture by Esther Weltevrede
As a follow up on Martin’s talk, I am going to present some empirical works. These project concern comparing engine results and customization of location. The aim of this study is:

1) Building on Googlization theory and search engine critique.
2) Empirical study. Reinterpreting Google for localization studies.

The key question is: What type of globalization machine is Google?
In this light, a couple of cases will be presented. Weltevrede starts by posing that PageRank is Googles way into the information environment. In an article published in 1998/1999 (?) PageRank is mentioned as the global ranking system for all pages, specifically designed for all the info of the world. Although Google states that they use a large number of variables, PageRank is primarily based on the link. The question of this research is: When Google moves to the local, what happens to the local results? What if we look at some services that are local:

A case:
Google “Amsterdam” and you get (respectively) red light, airport, coffee shops. This same query in Google.nl returns another set of results (arena, forest, tournament). Local domain Google is another method of localization (e.g. Google.de). There are 157 local Googles. The key variables are (as far as can be distilled: Google is not very transparent in providing this information):

• IP address
• top level domain
• webmasters page

If you visit one of these Googles (say, Google.be), you can also select pages from that locale (only provide me with result from Belgium, for instance). If you select this option, you get local results according to Google. Also, we notice that this particular Google is recommended in three languages (French, Dutch and German, in this case). Another way that Google returns local results is via ‘region” and of course a yellow-page kind of search is offered via Google maps. In answering what we can say about the type of machine Google is, Weltevrede states that it thinks globally and acts locally.

The first case study:
Local and international information sources. Research question: to what extend can the local domain Google present local results? Method used: query all the national Googles in their official languages. Then, in Google Translate, the search term is translated. The second step is to geo- locate sources. Instead of choosing for host, we looked at registration of the website. This is a more precise indication of who owns the website. The top ten results for the query “human rights”.

A picture is shows about the results. The selected national Google is Canada:

This map indicates that Canada has relatively many local results for ‘human rights’. We can also look at what the top results are globally. The UN is by far the largest source in the list. When we looked at the results more closely, the declaration of human rights keeps popping up. Often websites have translated the declaration in all languages they all call upon this source (On e can interpret this as a way of SEO) .

Next, a ranked tag cloud is shown.

We looked at these sources and blue- tagged sources contain the declaration of human rights. Next, a rank list of all countries queried is given. 40 % of all national Googles do not get any local results. If you look at the type list, you see that Europe leads the list, while at the lower end it is mostly African and Middle- Eastern countries. We can see that the local domain does not mean that you receive local information sources. How then are local results defined? Is it maybe language? A search is done on all Arabic countries. This shows a language web – a shared language space. Does that mean that there are no local sources? In Lebanon, the term “human rights” again is queried. While this does return results, these results do not make it to the top. Local sources are on the second page and beyond.

In order to test this claim (language) we looked at a geographical regions defined by languages: Europe is chosen due to its local and distinct languages. The visual below shows they have very local sources (Again indicated by black domain names). The EU- Googles hardly share sources – characterized by their local sources. This can be argued as a being a language web.

We now move to the last example: comparing two Portuguese speaking sources. Portugal compared to Brazil: Here we might conclude that the established local sources are privileged. Language webs prefer over local webs.

Search engine literacy (2nd case study)
We can use Google to read society; we have a particular way of interpreting search engine results. One example method: reading domain names and their main issues. Again, the example of human rights is used here. If we query this, we see a very established result list, where sources are relatively fixed. What happens when we query for a more dynamic topic? In this case a query is done on RFID in 2004;  back then, this was a young space. We see sources competing for the top. Compared to the human rights space, it has rather young and technology-minded sites; the build- up of the list is really different. Another method for research is to look at issues returned:

A third case study:
A query for “rights’ is performed. What is the top ten list of rights per country? The total list as shown. This research required reading and interpreting languages by the team members. The top ten of prominent rights in local domains were collected and visualized. The total image is shown.

The color code – blue rights are shared, while the black ones are culturally specific for domains.

If we zoom in, we see that in Italy, the unique rights are the right to forger and the right to nationality. In Japan, they have computer programming rights, for instance. In Australia, you have specifically man’s rights One favorite: Finland’s every-mans right to freedom to roam in nature. If we are to draw conclusion from this case study, they would be: the globalizing machine can show the shared as well as the locally specific. Google is localizing, regional, nation and local, showing shared and specific. Local results does not mean local sources. Also, different regions on the web are offered, mostly via language.

For more information, see Govcom.org and Digital Methods Initiative. DMI Project page on The Nationality of Issues. Repurposing Google for Internet Research.