Okay, so I know it’s naff, but for five years, the only way I’ve ever remembered Stevens’ On The Theory of Scale and Measurement is by the mnemonic NOIR:

• Nominal (named/categorised): apples, oranges, pears
• Ordinal (ordered): small, medium, large
• Interval (measured): the interval between 2:30pm and 3:00pm is the same as the interval between 10:30am and 11:00am
• Ratio (measured, has zero value): 0 tonnes of apples, 40 tonnes of apples.

What makes the NOIR concept important is that it can define every data type.

Then, you have Mackinlay’s Automating the Design of Graphical Presentations of Relational Information, which defines representational types, and can also be described by NOIR.

Put them together, and you have Zhang’s A Representational Analysis of Relational Displays which matches the data structure (NOIR) with the representation (also defined by NOIR).

What does this mean? It means the best way to represent a nominal dataset is by giving it a hue, as opposed to making different sizes of circles. Why? Because people might interpret, say, that blueberries are 10 times better than apples since area as a representation communicates a ratio data type.

So what does this mean for information aesthetic visualisation? Well, some have argued that form follows data and that’s the end of that. That data structure alone defines representation.

But data isn’t just a structure — there’s meaning behind a dataset, there’s insights, trends, feelings, issues that it addresses, people it involves… and THAT should all go into the representation as well.

So these are Andrea’s (pending) Information Aesthetic Meta-Representational Considerations:

• Feeling (leave people with an emotion): you can delight them, make them sad, or disgust them: just make them FEEL something. We’re not a soma-fied Brave New World.
• Interaction & Animation (give choice, show perspectives): it’s unlikely you can fit all the attributes of your data into one neat, 2D screen-based representation, so use interaction: this means we can add representations like animation (shaking, swooshing, scaling, flying, dropping, hovering…) and hovering, clicking, dragging, scaling, rotating…
• Learn (collaborate): information aesthetic visualisations should be collaborative. Just like ManyEyes, people should be able to leave comments, screenshots of their insights and things they’ve found. The meta-data is more valuable than the data in many ways, and should be shared and tapped into.
• Meaning (intent): your data (for information aesthetic visualisations, at least) is trying to tell you something! What is it? Is it trying to tell you that the environment is going to shit and we should do something about it? Is it trying to tell you that people living close to growers markets cook at home more?