A cou­ple of months ago, I came across a very insight­ful arti­cle with high rel­e­vance for infor­ma­tion visu­al­iza­tion: “More with less” in the always excel­lent ACM inter­ac­tions. It made me think quite a bit, and might also help some to under­stand a designer’s approach to visu­al­iza­tion a bit bet­ter, so here is the gist of the story (the fol­low­ing sec­tion mostly para­phrases the orig­i­nal article).

Propo­si­tional density

Let us start with the notion of a propo­si­tion: in this con­text, a propo­si­tion is sim­ply an ele­men­tary, atomic state­ment about the object at hand. “The FedEx logo­type is pur­ple” and “The FedEx logo­type is set in a sans-serif font” are propo­si­tions, and because they describe salient, per­cep­ti­ble prop­er­ties of the design, they are referred to as sur­face propo­si­tions.

Now, the FedEx logo became famous for a per­cep­tual trick: The white space between the E and the x cre­ates an arrow. This arrow induces, by its semi­otic read­ing, a num­ber of addi­tional asso­ci­a­tions and read­ings of the design: “FedEx is on the go”, “FedEx is forward-thinking”, etc. Note that these propo­si­tions, unlike the sur­face propo­si­tions, are much harder to enu­mer­ate as they depend on the mean­ing that the observer ascribes to the arrow. These are called deep propo­si­tions as they describe under­ly­ing and often hid­den mean­ings of the design. You can think of an ice­berg, where the sur­face propo­si­tions are over the water — easy to see and clear cut — but the much larger part is under water.

Now we have all ele­ments together to define propo­si­tional den­sity more precisely:

The propo­si­tional den­sity of an object is the num­ber of deep propo­si­tions divided by the num­ber of sur­face propo­si­tions it conveys.

Note that there are two ways of increas­ing propo­si­tional den­sity: Using fewer sur­face ele­ments to con­vey the same num­ber of deep propo­si­tions, or increas­ing the num­ber of deep propo­si­tions per sur­face ele­ment. Also, we can see now why the FedEx arrow is so effec­tive: It adds a high num­ber of deep propo­si­tions with­out adding sur­face ele­ments. It just emerges by clever arrange­ment of the sur­face ele­ments; in fact, any Gestalt law has a great poten­tial in this respect.

Gen­er­ally speak­ing, good design usu­ally has a high propo­si­tional den­sity. On the other hand, if your propo­si­tional den­sity is below one, you prob­a­bly have super­flu­ous, merely dec­o­ra­tive ele­ments in your design, which do not add to the deep reading.

Exam­ple: High altitude

We can use the image series “High Alti­tude” by Michael Naj­jar to illus­trate some of these con­cepts. For this series of pho­to­graphic mon­tages, the artist used self-shot pho­tos to assem­ble images of what seem to be nat­ural moun­tain scapes on first glance…

… yet the sky lines are care­fully crafted to match the devel­op­ment of major stocks and and indices. See for instance, the “dax_80-09″ image with an over­lay of the DAX index devel­op­ment I added:

Clearly, a high num­ber of sur­face fea­tures (and visual efforts) to express this sim­ple type of data (one mea­sure over time). Yet, I would claim that the high num­ber of deep propo­si­tions asso­ci­ated with this imagery jus­tify these efforts, and make the work an extremely rich type of visu­al­iza­tion; pos­si­ble asso­ci­a­tions include thin air, the bor­der­line between earth and sky, the man­i­fested his­tory of processes, mon­u­men­tal­ity, man-made vs. nat­ural objects etc.

Other works, like the “Global emis­sions” graphic pub­lished in GOOD mag­a­zine man­age to estab­lish a strong visual impres­sion suited to the topic with sim­ple visual means:

Finally, also a very sober visual style can yield rich expe­ri­ence by a clever com­bi­na­tion of chart style and con­tent; see for instance “Mir­a­cles in nature and Sci­ence”, a chart of the num­ber of men­tions of the word “mir­a­cle” in the renowned sci­en­tific jour­nals “Nature jour­nal” and “Sci­ence magazine”.

Note how the deep read­ing of the piece arises not only from the selec­tion of data sources (and thus the title pun), but also sub­tle design choices like the chart axis ratio.

What does this mean for visualization?

A com­mon tru­ism about infor­ma­tion visu­al­iza­tion is that it is pri­mar­ily about “show­ing the data”. I would like to chal­lenge this view. While this might be true for sci­en­tific (or finan­cial, or many other) appli­ca­tion fields, there are many good uses of visu­al­iza­tion that go beyond a pre­cise, “neu­tral” dis­play of data. In fact, most peo­ple I know are much less inter­ested in data than they are in infor­ma­tion. Infor­ma­tion can be con­veyed in many ways, and we have a rich tra­di­tion and cul­tural con­texts that allows us to com­mu­ni­cate beyond the sur­face level. For instance, in lin­guis­tics, there is a whole sub­field called prag­mat­ics that inves­ti­gates what is said beyond the “sur­face mean­ing” of the sen­tence, based on the con­text, con­no­ta­tions, the inferred intent of the speaker etc.

Bot­tom line

Every designed object invokes mul­ti­ple lay­ers of inter­pre­ta­tions and asso­ci­a­tions. It is not pos­si­ble to pro­duce a neu­tral design. In other words, there are always deep propo­si­tions involved. The shapes, the col­ors, the pre­sen­ta­tion form you use will always invoke asso­ci­a­tions. This has noth­ing to do with art, or “dress­ing up” data — the ques­tion is if you ignore this fact (and thus leave it to chance, if the deep propo­si­tions serve or con­tra­dict the intended read­ing) or acknowl­edge their exis­tence and use them con­sciously. In my view, tra­di­tional info­vis research has explored the sur­face propo­si­tions quite well — the next big chal­lenge is to under­stand the asso­cia­tive, evoca­tive char­ac­ter of our visual infor­ma­tion vocab­u­lary and use it effectively.

Adden­dum: While I have been pon­der­ing and edit­ing this arti­cle, a first empir­i­cal analy­sis of implied dynam­ics in visu­al­iza­tion has been pub­lished. Great!