Comment by saintx | Hacker News Reader

Comment by saintx | Hacker News Reader

Seeing people reason about this subject in the main article and the subsequent comments brings me back to my college "literary theory" courses. Instead of tossing up pithy declarative statements like "The best X is no X", or "The best Y is as little Y as possible", and then commenting on how these statements strike our intuition, we can bypass the literary methods in favor of scientific ones.

There's an entire body of research on visual design (a great introduction to which is Colin Ware's "Information Visualization") that gets right down to the nitty gritty details of human perception, like eye tracking speed, visual channel capacity, move and scan loops, visual working memory, the relationship between luminance channels, motion and shape as opposed to color channels, and on and on and on.

It turns out that "good" design principles can be quantified, communicated, taught, and learned, are based on our common biology, and that they needn't come at the expense of aesthetics or sound reasoning.

We have methods to test UI's to validate or invalidate them, we do not have a way to create new UI's from testing. HCI research provides a good grounding for design architecture options, but it does not dictate an actual design for you.

This is a very important distinction most test-only UX designers miss.

For example: If you test a skeuomorphic design versus a flat design, the testing will only tell you which of these interpretations of each design philosophy works best. However, maybe a better, more inspired, skeuomorphic design would possible perform better than flat if given a a different interpretation.

tl;dr -- testing only shows you which implementation works better in a set, not the best implementation possible.

And even worse only show you which implementation works in the context its being showed.

I.e. a lot of people UX people test was tested in pseudo environments.

http://000fff.org/getting-to-the-customer-why-everything-you...

Many UX experts laughed when this was written in 2010. But it is now the default IMO.

Testing has to be done in its actual environment to provide really useful feedback or be something completely (like a new way of interacting with your computer) new in order to be tested outside its final environment.

Testing no-ui is even harder since you have to build a "persona" before you can test the results.

Once you have an agreed goal, you can investigate techniques for reaching it. But some of the disagreements are over what the goal should be.

The "invisible interface" people's goal (in strong form) is to dispense with explicit interfaces as much as possible, seeing them as just friction between what people want and what the computer can do. One version of how to achieve that vision is AI-heavy, relying on people saying what they want in natural language and software like Siri figuring out how to give it to them.

The rough opposite is more of an interface-centric approach, where the goal is to actually foreground the interface as an explicit point of interaction, and have it present a logical and coherent set of controls that, rather than being invisible, makes clear "I am an interface and I do X/Y/Z".

Detailed information about things such as eye tracking speed seems to me like it'd be most useful in doing the actual implementation, once you have a rough idea of its purpose and use case. It's certainly possible to do scientific investigation of the more general question as well, but it requires more than micro-scale laboratory tests.

I agree with the literary analogy. This mostly seems to disagree with a classification that barely exists, or is still being worked on.

Reply on news.ycombinator.com