This post is the third of a five part series on ‘using the web for documentaries‘, addressing the following points: the embeddedness of society in the internet, the political in the web, the politics of engines, the politics of tools, and the web as an anticipatory medium.

Now that we have discussed researching the political in the web, let us have a look at the politics of engines to illustrate the need for medium specific methods on the web.

Before Google, search engines would rank paying customers higher. In those days search engines were not too good anyway, but boosting paid results would lead users to a lot of mischief. Then came Google, which recognized hyperlinks as an act of association. Their page rank algorithm measured the relative importance of web pages within a set of hyperlinked documents. Or in simple terms: if a webpage received a lot of hyperlinks it was bound to be important, given that the pages linking to that page were also important (in terms of the number of hyperlinks received). Google thus took an egalitarian approach to ordering and ranking the web and promised never to boost paid results. Their company motto became “don’t be evil”. But in the end Google wanted to expand to China, a totalitarian regime with a lot of censorship. To enter the Chinese market Google had to censor its search results by altering the results’ rankings.

Have a look at this url: http://www.langreiter.com/exec/google-vs-google.html?q=hungary. This service queries google.com and google.cn, comparing the results. Each dot is a search result; a blue line shows the same search result in google.com and google.cn Hungary is a quite neutral query for the Chinese, so you see many results being equal and and many of those with a similar rank. When http://www.langreiter.com/exec/google-vs-google.html?q=tiananmen is queried, however, you immediately see that the results are completely different and that those who are equal have a completely different ranking. This clearly is an example of politics built into the Google engine.

Let us now have a look at http://www.nytimes.com/imagepages/2006/02/12/weekinreview/20060212kahn_graph.html:

This research was carried out by the New York Times in 2006; it clearly shows the difference in image results for the same query, Tiananmen, in the American and Chinese versions of Google’s image search. The Chinese version only shows peaceful pictures of the square which in the whole Western world is known for its massacre by the Chinese government on Chinese citizens. The American Google version shows the iconic image of the individual who stood up against a whole platoon of tanks. Note, however, that the results of both Google Image search versions depict a strong ideology.

Let us recapitulate for a moment: we started off by stating that internet reflects society. Sometimes, too, traditional political research can be pursued when the medium is digitalized. Often, however, the political gets built into engines. It is also important to understand that the internet is not one single medium and that each medium has its own format with its own different affordances. Think for example about how the following has been digitalized, changed, or started on the internet: newspapers (e.g. Google News), encyclopedia (e.g. Wikipedia), pictures (e.g. Flickr), instant messaging for a quick chat, Twitter for status updates, etcetera.

The internet has changed the amount of data and the access to it. Engines typically aggregate content of a specific medium (pictures, news, blogs, tweets, …) and they rank and privilege certain content. It is also important to know that all engines do it in a different way: they all have their own way of aggregating, ranking, and representing; they all have their own politics: some rank by paid results, some by inlinks, some by numbers of sources, some by the amount of attention, some by date.. The questions then becomes: “how stable are the results coming from this tool used by everybody?” In most countries with a high internet penetration, Google is the leading search engine, through which most people access internet (in Holland more than 90% of the population uses Google as their default search engine).

In a research subject called Google studies, we wondered how steady Google’s results were for a particular query; so we made a tool called the Issue Dramaturg. This tool queries Google each day on specific phrases. Google in turn gives us a maximum of 1000 results per query, which we store in a database. We then look at the fluctuation in the ranking of websites for the queries we monitor. One day, we noticed that 911truth.org, an umbrella site for skeptics of the official 9/11 explanation, did not figure any more in the results for the 9/11 query, when before 911truth.org always snugly figured in the top 10 returns for that query. With the Issue Dramaturg, we were able to document the disappearance from Google of an entire website. Two weeks later, 911truth.org was back in the rankings.

Although there are different explanations about the disappearance (linking policy of 9/11 or plain conspiracy), it is clear that Google has its own built in politics. But Google is not our main point here, the main point is that each piece of software has its own built in politics, it enables certain actions and limits others.

By now, it is clear that each engine does something, that each engine has its own set of rules. However, each engine also works on a particular medium. An indexing mechanism (for example a search engine) claiming to give access to all available digital content of a particular medium, generates what we at DMI call a sphere. Google News for example may be considered to give access to the news sphere.

To shed light on the difference of spheres, let us have a look at the DMI research concerning animals in relation to climate change. Firstly, a list of endangered animals was compiled on the basis of the WWF, Greenpeace and LiveScience web sites. Secondly, we added the cow as a popular animal in relation to climate change (because of its methane production, a greenhouse gas). Then, different engines aggregating different media were queried for all these animals.

The following infographic shows a tag cloud of animals associated with climate change in the newssphere. The names of the animals are scaled in proportion to the number of results for the query ‘animal + “climate change”’ in Google News, i.e. their occurrence in the news sphere. It is clear the news made the polar bear into an icon.

When the same queries are input to Google, i.e. the web in general or the websphere, the results are more egalitarian: no particular animal stands out.

Looking further into the occurrence of animals in a particular sphere, DMI searched for the imagery of animals in relation to climate change. The following infographic shows images, scaled like a tag cloud, for the news sphere. Striking in this research was the professionality of the content and approach in this sphere.

The same queries for images in the blogosphere resulted in a lot of amateur photographs. The endangered animals and the cow are no top returns anymore, but the activists dressing up like endangered animals and the activist pets are.

Different engines aggregate different types of content (other kinds of media) and favor different kinds of information. This again demonstrates the importance of a medium specific approach . But even engines working in the same sphere, or on the same medium, probably will not show the same results for the same query. This easily becomes clear by looking at the following infographic . It shows the results for the same query (mastersofmedia.hum.uva.nl) in three different engines (Google Blogsearch, Technorati and Google). Only 5% results overlap in all engines, and only 19% overlaps in the engines specifically tailored for the blogosphere. Even in the same information space, using the same medium, the engines often deliver different results. This may be attributed to ranking mechanisms, but also to how and what the engine aggregates.

The previous examples clearly showed the built-in politics of engines: there are specific rankings, specific media on which the engines work and different kinds of source sets, amongst other things.

Apart from engines with built-in politics we are not always aware of, tools can also be devised for a specific kind of politics.

Next: the politics of tools