Trivialization for the Masses

There are thousands of visualizations on Many Eyes, but there is little in terms of further analysis and deeper discussion. There are dozens of visualization websites now that let you upload your data, but they all provide the same few visualization techniques and practically no analysis tools. While visualization for the masses may be here, we're not actually seeing much analysis from those same masses.

This issue came up last week at the Visual Analytics Consortium Meeting, in a panel that talked about responses global crises. While it's fairly easy to get people to whip up websites to collect data, there's practically no analysis of the material that gets collected.

A lot of open data is also like this: people go to great lengths to scrape data from websites and PDFs, but they don't seem to do anything with the results. The mechanical side of this is getting a lot more attention than the creative, analytical side.

On a different panel, Irene Ros also talked about Many Bills, and how the most popular view is one where you can't actually read anything, but see a colorful rendition of the bills.

A similar issue has bugged me for a while about open-source software, as well: most of it is either blatant copies of commercial software (OpenOffice, etc.), or infrastructure-type stuff (Linux, GNU, etc.). There is very little actual innovation in open-source software, most of it is reinventing or coding things that are well-known and clearly defined. Just look at the number of open-source text editors out there.

I think there is a pattern here. While a lot of people seem to be willing to do straight-forward, relatively mechanical tasks that are clearly defined and have an obvious measure for success (you've scraped all the data, implemented all the usual functions, etc.), there seem to be few who are willing to invest time in open-ended, ill-defined tasks that require a lot more creativity and independent thinking.

Now the question is, how do we get people to do the more interesting things? Just giving them interesting tools doesn't seem to do it. And neither does providing clever ways of interaction, like Many Eyes's live thumbnails. Is it just human nature to look for simple things to do when there is no real reward? Or can we figure out a way to engage people so that they look beyond the surface, however colorful and pretty it may be?

Posted by Robert Kosara on September 7, 2010.

---

Comments

Hadley Wickham says…

For the opposite, see some of the examples in the ggplot2 case study competition, http://github.com/hadley/ggplot2/wiki. There's some neat examples of using visualisation as part of a deep analytic process.

Robert Kosara says…

You're giving away a 32GB iPod touch! There is no question that there are good examples when you look at competitions, conferences, etc. But when there is no incentive other than semi-anonymous credit and warm fuzzy feelings, people will put in hours doing fairly mindless stuff, but not spend the same number of hours thinking deeper and figuring out.

Jock Mackinlay says…

Many Eyes only gave us a small number of view types. I am just starting to analyze the views published on Tableau Public. Many workbooks contain dashboards that combine multiple views into rich analytical experiences. The average is 2.2 sheets per dashboard with a tail up to one dashboard with 15 sheets. Almost half the dashboards contain filter widgets, which suggests an analytical experience involving raw data. I still need to collect data about brushing and linking. You avoid trivialization by making it easy to do open-ended authoring.

Joel says…

There is very little actual innovation in open-source software

Don't think I can agree at all with what you've said here, though that being said this might be a reflection more on the people who touch on the different areas than anything else.

First, while you refer to lack of innovation and point to applications that serve real purposes, one that you highlight (Open Office) is written off yet you didn't write off Microsoft Office for simply repeating what others were doing either?

There is nothing new - just incremental improvements. Some increments may be larger but at the end of the day it's all improvements on top of other improvements.

Linux, Apache, MySQL (and to a lesser degree Java) have all been some of the biggest game changers - leading to the development of the web. Wordpress as a tool is a major Open Source initiative, not to mention the open sourcing of application frameworks and languages - allowing for new innovation on the shoulders of other improvements.

Anonymous says…

Perhaps the reasoning is clearer if you look at it from the point of view of incentive: Incentive for mindless stuff like scraping, visualization mash-up etc is: a) Technical learnings & learning the tools of the trade (eg python, protovis, etc..) This is especially true for folks like me who are doing this as a hobby.

However, Incentive for thinking deeper, mining below the top-level visualization is low unless you are employed in that capacity or are a policy maker.

We need to find middle ground here. Any thoughts ?

Robert Kosara says…

The incentive would be to come up with exciting new insights. Learning a programming language or a tool is one thing, but figuring out something entirely new should be much more rewarding! I think the problem is the open-endedness of the task: you don't know where it'll lead you, and you might not even know when you're done. That seems to scare away a lot of people.

Toni Verbeiren says…

It all comes down to the same thing: too much rational, sequential, left-brain 'production' is going on. Too little parallel, broad, creative work.

Data analysis, visualization, presentation should be about delivering a message that sticks. In practice, little do we know about or invest in exactly this: making it stick. Rather, we focus on production of graphs in massive amounts, alternatives of expensive software, etc.

Sure, there are exceptions, as mentioned in the comments above. The difficulty, in my opinion, with e.g. data visualization is that it requires both analytical and emotional/intuitive insights. It's the only way to create something worthwhile.

To put it in the words of D. Pink, we need a whole new mind.

Hadley Wickham says…

You really think that people put together these examples just for an ipod touch?! I really doubt it - these are visualisations that people have created for their own work and their own understanding, and the prizes are just an incentive to get them to share.

OTOH I agree that crowd sourced analysis is basically ineffective - to really contribute to a data analysis, you need to immerse yourself in the data with a rich analytic toolkit. Manyeyes is fun, but gives nowhere near the depth you need to discover much other than the trivial.

Hadley Wickham says…

And lots of people are using visualisation to do so - but they're not doing it on open websites that provide shallow tools.

Danny Kingsley says…

The problem might simply be that the type of people who have the skill sets to create visualisation tools are not necessarily the type of people who have the abstract big thinking analytical minds. I work in the area of open access to scholarly publishing and am on the thinking side of the equation. It is very challenging because I am reliant on the technical people to build things for me and often there is considerable resistance because they don't always understand the point (or why precious resources should be diverted in this direction).

Matt says…

The amount of effort to interpret the data is much greater than just visualizing it. For example, you can create a scatter plot easily, but to evaluate whether there is correlation and causation is much more involved. As H. Wickham pointed out, interpretation requires you to "immerse" yourself in the data. You have to be an expert on the data to begin to interpret it. There may be interpretation going on, but not in public. Also, the interpretation is open to debate and people may not want to make the effort to take a stand on one side or the other. They're more interested in the tools and techniques.

Erik Anderson says…

Is it really that people want more expressive or "deeper" tools? There are plenty of websites that provide non-trivial visualization and even analytical capabilities. I think a lot of the problem is that people who are interested in combining analysis with visualization are doing so for research purposes. If you do that in a public forum you run the risk of killing a publication before it really gets off the ground. Combined with the "publish or die" mentality at research groups around the world, it provides incentive to hide your results and analysis as much as you can.

Chris Pudney says…

Robert, your lament is echoed elsewhere in a recent blog post by Stephen Few regarding business intelligence software. The wall Few refers to separates "technology-centric, engineering-oriented" activities from those that are "human-centric, design-oriented". I see visualization being the former, and analysis the latter.

Hadley Wickham says…

I don't think the fear of being scooped is the biggest factor - it does take more effort to work openly, especially when web based tools for data analysis are not as powerful as desktop based. Why work on your analysis publicly when it's much easier to do it privately?

Tobias says…

I think many people respond well to challenges. The reward can be monetary such as the famous Netflix prize, but more often than not - solving a problem and presenting a clever solution is enough of a reward. The ggplot case study competition that Hadley Wickham referenced in his first comment to your post is a good example. The bi-annual data expo that the American Statistical Association runs or the VAST challenge at visweek are other good challenges that get the creative juices flowing.

It would be interesting to see a site that combined tools for visualization and analysis with the ability for others to ask questions and issue data related challenges. Think of Many Eyes meets Yahoo Answers or Stack Overflow.

Aneesh Karve says…

This RSA Animate talk suggests that, for non-mechanical tasks that require creativity, Autonomy, Master and Purpose are the key motivators. The talk further argues that reward detracts from performance in creative tasks.

On a diagonally related note, Gladwell's Outliers cites Autonomy, Complexity and Reward commensurate with effort as the pillars of meaningful work.

Martin Theus says…

There is certainly a difference between manyeyes (which is mainly a "visual dump" of data, which hardly can tell us the story behind the data) and the ggplot case studies (which show the presentation graphics of what was the result of an analysis).

But in both cases we do not see anything about using graphics for actually exploring the data. This is not really surprising, as the cumbersome process of analyzing data (which might not be too cumbersome with the right tools) often goes into dead ends and several iterations, that no one really wants to present.

When we want to promote the (right) use of graphics, we need to go beyond showing the polished results in fascinating graphics - this is what people know at least since Minard ... The real challenge is to convince people that their analyses can benefit greatly from using graphics. Most statistical and visualization packages fail to support this kind of thinking - at least they don't encourage the user to use graphics in exploration.

Venkat says…

I agree to this PoV. Either we see too much InforGraphics which are , for the lack of a better term, 'posters' or see charts being thrown all over the place.

Efficient Slice-and-dicing of data should be provided so that interpretation-cycles are faster. 'Visualization' is one of the key factors in reducing this interpretation cycle, as the user shouldnt be made to sweat when he analyses the data.