Tag Archives: Bernhard Rieder

Teaching alternatives ways of searching

I was reading a draft of a paper for Bernard about democritizing web searches.  I won’t go into the details of the paper–it’s not published yet.  But, I do want to mention two search tools that I had never heard about before reading this paper, and I’ll give some short quotes from Bernhard on those:

  • Clusty — a search engine that divides results into thematic clusters that users can use to navigate up to 500 results at a time. The cluster list provides a first overview over the search subject and by showing aspects that users were not aware of direct them in new directions.
  • TermCloud Search — a search interface designed to map a topic rather than provide the shortest way between a query and a document. Using the simple tagcloud principle – keywords are shown in different sizes according to relevancy – the goal is to make the user aware of the concepts surrounding her query and to encourage exploration rather than quick answers.

Both of these could be really useful in teaching students about searching because they offer an alternative to Google’s approach of trying to give users the most relevant (meaning most popular) results first.  I think these could alleviate the concern many teachers and librarians have about students who just google everything rather than going into the library and browsing shelves–that they don’t make the same connections or experience the serendipity that can arise from looking around rather than right at the result you want.

Bernhard Rieder and Algorithmic Proximity at IR9

The last talk I saw that I’ll report was Bernhard’s, on Algorithmic Proximity. Bernhard started off with background on the work he and Mirko have done that led up to the hybrid foam model, but his main question in this talk was to look at lower level sociality, such as in sites like Flickr, where most interactions are singular, and connections are fleeting. He is trying to understand “socio-genesis” or the process through which these low level communications crystallize into a real relationship.

In reality, individuals stand at varying social distances, or in network theory terms, where individuals are linked by paths of varying lengths which represent the probability of association. Add to this the notion of homophily; that we tend to associate with those like ourselves. (on the twitter channel for IR9 a number of people agreed that while it was true, we hated to admit it because it seemed narrow-minded).

Next, it is possible to render social interactions digitally and what will that reveal? Skipping the math… we see the importance of space somewhat reduced, and status homophily seems to be replaced by value homophily, where interest factors become more important than socio-economic factors.

Algorithmic proximity is a form of social proximity produced by the rendering of many factors in order to make recommendations about friends or matches. For example, on Facebook, the number of friends you have in common with someone may lead to a friend recommendation in “people you might know.” This is most noticeable on dating sites which aim to match people based on similarity across a range of categories, and in fact is almost essential if one is to effectively filter through all the possible matches. Bernhard went through a few other examples; Last.fm, Flickr, and Delicious, and said a bit about how on these sites, similar tagging practices might lead people to start following other users.

But what about serendipity? Is homophily a feature or a bug? If we only see people who are like us, then what? I think that’s a frightening prospect myself; I can think of a lot of interesting ideas and people I would hate to have missed, but if all my encounters were based on some kind of homophily, we would never have met. A fun counter example, the Unsuggester. This site tells you what books you would hate based on books you like (and maybe by extension, the people). I’m afraid I do judge people by what they read, sometimes….

I really need to get the whole paper because I think the math would be interesting, and also, Bernhard makes very strong but closely argued points, and a lot of the details have to be left out of such a short talk. So I’ve emailed Bernhard and if I can get more details, I’ll update this entry later because this seems important to me, thought it’s a tangent to Bernhard’s work.

If I am to figure out how people connect and stay connected, I think this could be a really important piece of the puzzle, and also suggests measurable data I could look at in order to see patterns — for example, what kind of proximity, exactly, seems most important? Are there certan values or other shared chracteristics that correlate more strongly with connection than others?

A really thought-provoking talk.

My Panel

I don’t want to brag..well, actually I do. The panel went very well considering how many speakers we ended up with. Everyone kept to the time limit, no one had technical problems. And the talks themselves were all quite good; I think even exceptional in going beyond the anecdotal case studies we so often see when it comes to work on participation. Since we had so many speakers, there was really no time for discussion; that was the one downside, but I did have some short chats with people later on about our panel, so I guess they liked it.

Here is a link to my prior post which has links to all the full papers.

I also recorded audio for the whole panel and hope to eventually make podcasts for each speaker.

Big thanks to Elfi, Anders, Christian, and Mirko. You guys rock! 🙂

Finally, I really have to thank Bernhard Rieder for his masterful work as respondent. He had quite a job having to read all five papers and find some way of summing them all up. I also recorded that, thankfully because Bernhard had good ideas that inspire further development of my ideas at least. –I heard the same from Elfi, in fact.