Data Art vs. Data Visualization: Why Does a Distinction Matter?
Two distinct approaches to presenting data graphically exist today—data visualization and data art—and rarely do the twain meet. They differ in purpose and in design. When we fail to distinguish them from one another, we not only create confusion, but do great harm as well.
There are as many definitions of data visualization as there are definers, but at the root of this term that has been around for many years is the goal that data be visualized in a way that leads to understanding. Whatever else it does, it must inform. If we accept this as fundamental to the definition of data visualization, we can judge the merits of any example above all else on how clearly, thoroughly, and accurately it enlightens.
By data art, I’m referring to visualizations of data that seek primarily to entertain or produce an aesthetic experience. It is art that is based on data. As such, we can judge its merits as we do art in general.
Either one, done well, is worthwhile, assuming that it fits the task at hand. If the task is to help a particular group of people understand something, then data art is not appropriate, no matter how well it is executed. If the task is to entertain or engage an audience in a particular emotional experience, then data visualization probably isn’t appropriate. If the situation requires that both objectives are achieved, then a deeply informing and aesthetically beautiful visualization would be in order. Although it is quite easy to make any data visualization aesthetically pleasing, it takes a great deal of skill as a visual designer and information communicator to make one beautiful.
People make better decisions when they’re based on understanding. For information to be understood, it must often be presented in visual form. This is because patterns, trends, outliers, and a sense of the whole as opposed to its parts require a picture for the human brain to see and comprehend. Data visualization is essential. Visualizing data effectively is vital. Anything less is frivolous, costly, and harmful.
How in particular is data art—visualizations that strive to entertain or to create aesthetic experiences with little concern for informing—harmful when it masquerades as data visualization?
- It suggests that data cannot be visualized without training in the graphic arts. As such, it works against the democratization of data. In fact, anyone of reasonable intelligence and a little training can present data effectively. It’s vital that this ability spreads more broadly across the population, because it can play a role in making a better world.
- It features ineffective practices as exemplars of data visualization. It encourages people to present data in ways that are difficult to perceive and understand simply because they are prettier or more entertaining, which is rarely relevant to the task.
- It keeps the practice of data visualization spinning its wheels, never able to progress beyond the mistakes of the past. Best practices of data visualization have emerged through many years of research and experience. “Those who cannot remember the past are condemned to repeat it” (Santayana).
I am personally and painfully acquainted with each of these problems. For this reason, I try to differentiate data art from data visualization and encourage others to do so as well.
Take care,
22 Comments on “Data Art vs. Data Visualization: Why Does a Distinction Matter?”
Well said. I think it’s important for the creator (whether analyst or artist) to make their intentions clear; so their work can be appropriately evaluated.
If they would admit their work is data art only, they’d be likely to receive only subjective criticism. Yet so often they try to pass it off as data visualization instead, and are surprised (and offended) when you feature it here as what not to do.
Maybe it’s because so many of them are out to make money, and a disclaimer saying, “Warning: This diagram is for artistic purposes only, and is not recommended for anyone who wishes to understand the data,” might lessen their profits. You’d think that they’d just learn how to present data in an informative way so it wouldn’t be an issue. But I suppose it’s hard to justify learning something when they (and their followers) already consider themselves to be visionaries.
Nevertheless, I’ve learned to mentally attach a disclaimer like the one above whenever I see a fancy-schmancy infographic anymore.
My position is exactly that of Steve’s. As Steve said, there are as many definitions of data visualization as there are definers. Many of these definitions include data art.Therefore, when I submitted a proposal to a conference on this topic recently, I used the title “Data Visualization: Statistical Graphics or Data Art.” Statistical graphics does not have the multitude of meanings that data visualization does.
I recently attended a session on this topic presented by Leaders in Software and Art and the New York Public Library Labs. In the session, Andrew Gelman proposed a compromise for situations where it might be appropriate to use a figure that grabs attention. It is the click-through compromise. Data art is shown to attract attention. The reader then clicks through and gets a statistical graphic that facilitates understanding. Clicking through again gives the raw data.
This is not necessary for decision making in business environments, but may be a useful compromise for the media.
Naomi,
Andrew Gelman presents this opinion more thoroughly in an article that he co-authored with Antony Unwin for the Journal of Statistical Computing and Graphics titled “Infovis and Statistical Graphics: Different Goals, Different Looks.” I was asked to write a response to this article, which will appear soon in the same journal. The notion that a member of the general public must be enticed to look at statistical data through the use of a chart that doesn’t effectively inform but is somehow artistic, which can then be supplemented by an informative statistical graphic and a table of numbers, is an unnecessary and ultimately ineffective compromise. It is based on a false assumption that people will not care about information unless it is presented in an aesthetically stunning manner. A good title that catches the reader’s interest can provide the only hook that’s necessary to catch a reader’s interest. Gelman also seems to think that a good statistical graphic must be drab, which certainly isn’t the case.
It is important to display data in aesthetically pleasing ways (who wants to look at an ugly chart?), but it is rarely necessary to employ the talents of a graphic designer to make a chart beautiful. A fine statistical graphic can be aesthetically pleasing in ways that don’t distort the information or distract from it. Gelman encourages statisticians to team up with graphic designers to get their messages across, but they could more easily and effectively engage and inform readers by learning a few simple principles of visual design. The skills that statisticians need to present data effectively are well within their reach, but few have taken the time to develop them. This is tragic.
Stephen – Excellent posting!
Data art for business purposes is like providing assembly instructions for a new PC in haiku. It may be beautiful but it is a barrier to communication.
You give three persuasive arguments against data art that tries to pass itself off as information. May I add a fourth?
Data art tends to make the data look more complicated than it is. As such it is off putting to people who are not confident with numbers and leaves them with the impression that maths and charts are beyond their abilities. For people who are confident with numbers, such art obstructs interpretation.
Further, like you say in your follow-up to Naomi, a good title can catch the readers’ interest more appropriately than artistic-add-ons. Yet the convention is for titles such as “imports of oil†rather than the more engaging (and more helpful) “Oil imports up!â€.
This is a matter of public education.
And an uphill task for us people involved in data presentation.
Excellent post.
The data art crowd often argues that people uncomfortable with statistical graphs and numbers are more likely to study a data art more than a data visualization. But, are there any studies or evidence that such people do get a better understanding of the material from a data art then by just reading the article or piece that the data art is illustrating? In my experience, data art are often confusing and may even distort the data. I.e., the opposite of what people not comfortable with numbers need. That group need well designed, simple visuals that communicate the data efficient more then experienced people skilled in visual display of statistical data.
Naomi mentions the “click-through compromise”. If data art is confusing and poorly communicate the data, should they be used as “attention grabbers”? Well, I really don’t see why one should confuse the reader before you give them the “real” visuals. A strong photo, a humorous cartoon or some interesting typography (there are many striking, classical examples) relevant to the case/story would serve as attention grabbers without distorting the data or confusing the reader.
@Anders: “If data art is confusing and poorly communicate the data, should they be used as ‘attention grabbers’?”
It could be that data art on its own bears so many unanswered questions that the viewer is forced to read the article if they want any information. Viewers forced to read the article spend more time on the page, and thus more time viewing ads, which could lead to increased ad revenue. The “click-through compromise” might be similarly-motivated: increased page views (by clicking through to the next page) means increased ad views. It seems there could be good money in portraying data “artfully”.
My speculation might be a bit cynical, but it’s really the only excuse I can devise for using data art to “grab attention” for what otherwise must be uninteresting information.
Very well said. Some of the worst Data Art has promoted itself as a means to address (and even solve) global social problems. In reality, such designs don’t address complex issues and instead deal in bias, bad research, and a misunderstanding of underlying data. Worse, organizations showcase such designs with much praise and no scrutiny.
Steve,
As usual, well said and unambiguous.
This discussion is particularly interesting to me, because my background and experience is at least 50% artsy and 50% hard core “analytics”. Still, I am always surprised when folks who do statistical analysis to inform strategic and tactical business decisions relegate visual data sensemaking to the “after the work is done” phase, “good for presentations – you know, take the answers and make some pretty graphs.” It dismays. Then I re-read your blog posts and remember …
Thanks.
MANY BLESSINGS!
Michael
In defense of Data Art:
I think this distinction is important, but this definition of Data Art is far too narrow and overly simplistic. Data visualization is analogous to graphic design. It’s about communicating a concept clearly (often with a purported sense of objectivity). Data Art is art. It’s about asking questions, creating experiences, telling stories – it is subjectivity.
To say that art is about creating an aesthetic experience is limiting and far too narrow. It’s about engaging the mind in a more exploratory manner and raising questions and discussions as opposed to answering questions.
While there are good points here (most of which I agree with), it reads like a scientist denying the legitimacy of communications and thought outside of the lab environment. I understand the point, specifically that the confusion between these two practices is problematic – and I agree. However, like da Vinci practiced science and art there is a healthy balance of creativity and analytical thought which leads to better communication and critical thinking.
Fundamentally I think my issue is one of terminology. Here I’m seeing the term Data Art as a catch-all for failed visualizations – specifically things that are purposefully confusing in an attempt to be beautiful. While this could be a form of art, I would argue that it’s not representative of the category – and even further that good Data Art also does not strive for this end. Instead it too attempts to beautifully engage the mind, but towards a different purpose – not one of answering a question, but not one void of meaning and purpose either.
Finally, I’ll make the argument that there does exist an overlap where Data Visualization and Data Art meet. Where a thought is conveyed incredibly clearly and accurately in a beautiful and engaging way – making a point and making one think, and this may in fact be an incredibly valid pursuit.
Aaron,
By titling your response “In Defense of Art†you have suggested that I am attacking art and diminishing its value, which is not the case. Life would be dull without art. I agree with almost everything that you’ve written and appreciate the quality of your thinking and communication.
My description of “data art†is intentionally incomplete. Art is difficult to define, and a comprehensive definition isn’t required for the point that I’m making. There is a distinction between graphical representations of data that are meant to elicit understanding and inform decisions and those that are meant to entertain, create an aesthetic experience, or anything else that art might seek to elicit. I am not using “data art†as a term of derision for failed data visualization. Whether a graphical representation qualifies as data visualization or data art is not a matter of effectiveness, but one of intention and design. Just as data visualizations can fail, and often do, because they are poorly designed, attempts to create art based on data can also fail through poor execution. Nevertheless, I’ve seen beautiful examples of data art that elevate me in ways that elude understanding. I have at times critiqued such examples harshly, however, because their creators insisted on calling them data visualizations, even though these works failed to reveal the meaning of the data.
I agree that data visualization may exhibit exceptional qualities of art. This is rarely achieved, however, and it requires a level of skill that few of us will ever possess. Even if we had these skills, most data visualizations are used for purposes that can’t wait for an exquisite fusion of form and function, which takes time. Most of us who rely on data visualization to achieve understanding and inform decisions don’t need the skills of an artist. To do its job, data visualization rarely needs to be beautiful or otherwise artistically evocative – it needs to inform.
I’ll leave it to the data artists to define “data art†in a comprehensive manner. It is different than data visualization and ought to be evaluated on different terms. A problem that we’ve facing today is that many organizations – government, corporations, non-profit groups, etc. – are looking for artists to help them use data in ways that don’t require art and they’re hiring people who have no training in data sensemaking. They are often spending a great deal of money for pretty pictures that fail to inform. This is a harmful waste.
very thought provoking post. Thank you.
Data Visualization to me is more Scientific and logical (guiding audience to think in a specific direction)
Data Art to me is creative way of expressing things (may be more of touching to heart and without channeling my of thoughts towards a specific direction)
Kiran
A key issue I have with data art, and a lot of data visualisations is that the excessive importance attached to the presentation leaves failings in terms of the validity of the data or information being presented, comparing like with unlike is a key issue in my experience. In my view, data visualisation done with an incomplete understanding of the underlying data is worthless, regardless of how beautiful it might be. As such, there’s a basic requirement common regardless of how artistic the work is to be when the base subject of a piece of work is data.
Today, I received an email from Tableau about their new “Viz of the Day,†featuring visualizations uploaded by bloggers and journalists to Tableau Public. I couldn’t help but recall this blog post as I read the description of their new endeavor, which I excerpt below:
—
Two years into the Tableau Public experiment, our authors have created thousands upon thousands of amazing interactive visualizations…today we are proud to announce Viz of the Day, a once daily feed of beautiful visual stories.
The concept is simple: every day, bloggers, journalists and data enthusiasts from Australia to Alameda to Argentina create hundreds of visualizations using Tableau Public…every weekday we will share a beautiful, amazing visual story from the Tableau Public library through Viz of the Day.
Which visualizations are we going to choose? It is impossible to pick just one archetype, but one thing is for certain: they will all be beautiful, meaningful, interactive and informative.
—
From: http://www.tableausoftware.com/about/blog/2012/05/delivering-visual-stories-through-viz-day-16741
Why must visual stories be “beautiful”? As I looked at their first ever Viz of the Day, I was not struck by its beauty. If anything, I was annoyed by the stray pie chart.
Jordan,
I share your concern about the degree to which Tableau Public is being presented as a tool for making “beautiful” data visualizations. I believe that this spin on the product is harmful; it certainly undermines the unique power of a well-designed data exploration, analysis, and presentation tool such as Tableau. This emphasis is too closely aligned with the perspective of graphic artists who care mostly about making infographics beautiful rather than the perspective of Tableau’s primary audience: people who actually work with data to promote understanding and the better decisions that result from understanding.
Stephen,
This blog post was cited in the article Data Visualization VS Infographics on the site http://www.dashboardinsight.com as support for the article. I was wondering if you read the article and what your thoughts on it were.
Scott,
Let’s begin with your opinion of this article. What do you think?
Thanks,
Steve
I agree with what the article says. A lot of people confuse Data Art(Infographs) with Data Visualization and therefore draw the conclusions that are stated in the article. This lack of distinction often leads to graphs that inform not even being created in favor of just supplying the raw data, which may sometimes be hard to pick out patterns in the data.
This is one of the few articles that has sited you in the correct way and not taken things out of context.
Scott,
My only concern with these comments on Dashboard Insight pertain to the last sentence: “Our hope is that industry leaders, from people like Stephen Few or Edward Tufte to Data Visualization providers like Qlikview, Dundas, and Tableau, will continue to educate the business world and shed this negative stigma forever.” Notice the inclusion of “Dundas” in the list of data visualization providers. By mentioning Dundas in association with my name and Tufte’s, Dashboard Insight is implying that Dundas follows the data visualization practices that we teach, which is far from the case. Dundas exhibits many tragically bad practices in their products. Dundas is not a major player in the space, so why was it mentioned? The answer is: Dundas owns Dashboard Insight. Dundas uses this Dashboard Insight, which presents itself as independent and objective, for its own marketing purposes, which is deceitful.
Scott,
One more question. You said that “this is one of the few articles that has cited you in the correct way and not taken things out of context.” I haven’t run across any articles that have cited me incorrectly. If you have links to any, please share them with me.
I don’t think it was anything recent. I will let you know when I do. Keep up the good work.
This nauseating video by the social marketing company, Column Five Media, seems to be ripped from Show Me The Numbers and Tapping the Power of Visual Perception: http://www.youtube.com/watch?v=xekEXM0Vonc. They don’t cite you specifically however, which may be a blessing in disguise since most of their infographics are awful.
Jordan,
None of the information that appears in this video, which also appears in my work, is an original product of my work. I learned derived this knowledge from others, such as Jacques Bertin and Colin Ware, and cited them as the sources wherever appropriate.