Data visualization for modeling, not only for storytelling

Many people say that storytelling is necessary for good data visualization. It is a one of the most popular word to praise a data-viz project.
I agree that it is necessary for sure, because if you see a data-vis lacking context and story behind, it could be a mere abstract art showing circles, lines, colors in order.

But I think the term and its point of view has been overused today. Data visualization can do more with other fundamental activity for human’s mind and communication.

Data visualization as a model of the reality:

What are models? Like models in physics or economics, a model is a simplified object from data in the real world. It also has a predetermined structure that covers data and divides data into some attributes. Because of the structure, the model has space to fill with data.
You can then add new data to it, but more important functionality is being able to infer new data by leveraging the structure. It can even join new kind of data set with it.

 

A good example of data visualization as a model is periodic table by Dmitri Mendeleev.

Periodic table

Periodic table

He made the model by aligning in the number of protons in atoms. It was the great idea that lining up the dataset  by properties of the elements in the periodic cycle, like calendar.

It’s not just a visualization of data. The table provided viewers with gaps to fill the blanks. It was very useful in the age when many rare atoms haven’t discovered yet.
I don’t know what story viewers can read from it. It is not an storytelling or art but a puzzle game as Sudoku.
So you can put new data which you think it  is supposed to sit on.

 

While Mendeleev’s table got the big success in history. The form is not needed to looks like that.
I found some different form of visualization of atoms on Wikipedia , 周期表 – Wikipedia.

The Ring Of Periodic Elements (TROPE) by Alexander Braun

The Ring Of Periodic Elements (TROPE) by Alexander Braun

Each visualization differ in shape and concept. But the most important fact is they have different attributes being able to extend.  Mendeleev’s one can be extend data by filling boxes of the calendar, but others have angle, stack, height, direction, or distortion.

And data that you would extend partly depends on  the metaphor you perceive.
I think  the second example is very like plant’s cell division on the tips of roots. Such analogy can drive you to search other data set in you mind. For example, what data should be adde when the root extrude more to outside? Or how is three dimensional form? Are they molecules? How do the atomic roots grow from the start of the universe to the end in amount?

 

Data visualization as model is a mind game:

So the models are a tool for mind game, like mind map. But doing mind game in data visualization has advantage other mind games don’t have.

Basically, mind games including mind map have week ties with reality. Almost all mind game is used to present, conjuncture, exchange image with participants. It is great for planning next business plan and organize intelligent to attack new area.
But for finding and questing new fact from data, the visual image should come from actual data.  Data visualization allows people to get reality in mind, modify it, and extend it.
It is an alternative reality that can be share.

Visual is good to share information in group. So data visualization is nice for communication tool through visual and data.
So the first easy step to play mind game in data visualization is asking someone or yourself that what does the chart looks like?  Then ask what data should be there if it is like something you said?

 

Interactive data visualization must allows user to extend data from original data:

From the view of advocate of model, today’s many “interactive data visualization” is wrong wording. Many of such allow users to pick up a fraction of data, and change form of a chart. The interactivity stays at presentation. Indeed, such specifications are graced by analyzers who are interesting in the subject very strongly, like stock traders or CIA intelligent officers. But for many ordinary people who have relatively  modest interest in a subject, close-up functionality is a over technology. We don’t care so much in detail.

Instead, true interactive data visualization should allow users to modify and extend data by change the chart’s property. In the chart, the original data is just a template and a use-case.
The real benefit to use it is that a user can present his mind in the form of visual, to other people.
I think the data and properties are not necessary to be actual data set. Users can make hypothetical data as a prediction.
Then, when users’ mind are gathered and visualized overlaying the original chart, it is time for bayesian. I mean the visualization itself is a visualization of collective intelligence growing from real data.

Possible challenges to implement it:

As every webapp that relys on user generated contents, demand of users is the key to success. People want to communicate with each other through data visualization? It’s utterly unclear.

And there is also the big problem. Combining different dataset is sometimes very hard. Making data model visualization that can invite other kind of dataset seems to be the hardest thing in implementing it.

Leave a Reply