IBM Many Eyes After One Month

ReadWriteWeb writes "IBM's Many Eyes app, a 'shared visualization and discovery' service, has been running for a month now. In this article two of the IBM researchers behind Many Eyes, Martin Wattenberg and Fernanda B. Viégas, showcase some of the best visualizations so far. They also talk about the future of 'social data analysis' on the Web. Wattenberg and Viégas believe that Many Eyes is not just social software, but 'societal-scale software.' They say that Many Eyes represents a break from conventional visualization research. Traditionally, computer scientists concentrate on scaling in terms of data, making visualizations work for bigger and bigger databases. IBM's agenda with Many Eyes is to scale the audience, not the data."

Comments

Too many demands on eyes

[Animats]

Too many people are trying to make others do work for them for free. There's only so much attention to go around. And we're running out.

Wikipedia made people think this could work, but Wikipedia today is mostly cruft. Most of the good articles were added when Wikipedia was a tenth the size it is now. What's coming in now is mostly dreck. Existing articles suffer from ongoing churn, as people make marginal edits and others revert them, without much real progress. Jimbo got out at the peak of the bubble.

Then there are all those "rating sites". Those suffer from a scaling problem - rating only works when the number of raters is large compared to the number of things to be rated. Otherwise, stuff gets rated up by people promoting it.

What we need is more automation, not more eyeballs.

Re:Too many demands on eyes

[mebollocks]

Sorry but this is just wrong.

"Too many people are trying to make others do work for them for free." Really? How so? How many is too many? Perhaps you think too many people are willing to 'work', as you call it, for free?

There's only so much attention to go around. And we're running out. Now we're running out of attention. Are you sure? Better get back to doing what we used to do and watch tv then so, before this silly attention-stealing, intarweb came along then so.

Most of the good articles were added when Wikipedia was a tenth the size it is now. What's coming in now is mostly dreck. Existing articles suffer from ongoing churn, as people make marginal edits and others revert them, without much real progress. Well all its "good articles" (the articles that interest most) are finished. Wikipedia never becomes finished because at all times it's a snapshot of how society and culture sees itself at one moment in time.

Then there are all those "rating sites". Those suffer from a scaling problem - rating only works when the number of raters is large compared to the number of things to be rated. Otherwise, stuff gets rated up by people promoting it. So we need more raters, "eybealls", if you will.

What we need is more automation, not more eyeballs. So we don't need more raters, we need some magic algorithms that can extrapolate the truth from fewer raters. hmm...

Re:Too many demands on eyes

[tezza]

1. There's only so much attention to go around.

I could not agree more. More specifically, there is only so much trained attention to go around. In the case of Many Eyes, interpreting the visualization takes a uni graduates equivalent of training. Not a degree, but similar capacity. Other slashdotters may argue to set the bar lower, but how much lower than HighSchool Grad can it be in educational terms? People boggle at the concept of Compound Interest, these mutli-layered datasets would be hard for them. E.g. Some of the data needs to be on a logarithmic scale, which is complicated to grok.

2. The Big Picture often misses The Little Picture

Often with these overviews, important details can be overlooked.

a)Metrics not chosen well: For instance if the visualization has been weighted to represent one metric you can entriely miss the metric you should be looking for.
b) Too many entries: A classic case in point is Wikipedia. If you stumble upon a complex Disambiguation you are lost by the number of entries. This is the problem that Google set out to minimize. It has improved on a couple of years ago but all the Link Farms still hamper the same problem of finding what you are looking for.