Hacker News Reader

Comment by techiferous | Hacker News Reader

I'm a big fan of minimal interfaces. But I guess minimal interfaces are good for the web user who is looking at your web app for two seconds before deciding whether to keep looking or move on.

This actually looks like a pretty effective interface for users who are willing to be trained for months. Think about it: after you are trained, every piece of information is just one glance away and every action is just one button click away.

I'd love to hear an opinion from an expert, though. sinks back into his armchair

View on HN

At my institute we had two electron microscopes. One was 20 years old and every setting had it's own button or knob. It looked like a space ship. The other one was very new and only had a joystick plus some terrible software.

People _loved_ the old one because having lots of buttons you can tune almost in parallel was much more efficient than the modal click-here, use-joystick workflow.

The newer EM generated much better images though, so the choice wasn't just between UIs.

Usability considerations play a large role when designing complex interfaces. Complexity pretty much has to be reduced and errors have to be displayed in a way that makes it possible to actually deal with them, especially in emergencies. Everything just lighting up may be nice for situations when there’s time to investigate but certainly not when time is critical. Yet wrong decisions become more likely when all the facts in all detail are not known so there is also a right balance to strike.

There is an interesting talk which explores UI problems in airplanes (not quite Space Shuttles but also with a lot of buttons and, as far as the UI is concerned, with a lot of similar elements): http://www.youtube.com/watch?v=YlEcCfEakro

Reminds me of the often-told story about Apollo 12 and how a little-known switch resolved an electrical issue that threatened the entire mission, though in that case it was a person on the ground who instructed an astronaut what to do. The following four-minute clip is from the documentary "Failure Is Not An Option."

http://www.youtube.com/watch?v=eWQIryll8y8

I found a picture of an A380 cockpit and it appears to have that design philosophy. http://upload.wikimedia.org/wikipedia/commons/a/a4/Airbus_A3...

Nuke plants have a great many buttons, dials, gauges and levers, too, according to one of Don Norman's books.

ADDED. Norman said that one of the advantages is that when one operator lowers the aperture on value 68 (or whatever) the other operators can easily tell that that is what he is doing.

As you can see there are a lot of big screens and very few actual buttons and switches in that photo. It looks like the four keyboards are the bulk of the buttons. I would be curious to know how modal those screens are and how much information they can display at one point in time. It doesn’t look like they can display much and I can’t imagine that it’s fun to go hunting for necessary information in an emergency. Depending on how that works it might be a usability nightmare, one can only hope that Airbus does plenty of UI testing.

(In that context an interesting question is what this interface is optimized for. It may work great on normal flights – that’s nearly all of them, by the way – but break down horribly in emergencies. I guess the UI design rule of designing for the common and not the rare case doesn’t really apply here.)

Actually I was reading an article (I can't find it now) about that Qantas A380 engine explosion, and it said that this was a genuine problem in that situation. Once the engine blew up, there were dozens and dozens of different errors popping up, and the first thing the pilots had to do was to page through all the error codes and prioritize them to figure out what the real problems were.

In many ways, better to have a single giant control panel which lights up a different light for each error that can occur.

This is actually an artifact of the way planes are certified today (DO-178B). The verification process has to be done by hand - you can't verify automatically generated computer code.

The result is that you can't build say, a prolog program which produces a huge switch statement to prioritize errors. As a result, there effectively is no prioritization of errors and warnings at all. They're all treated as the same severity.

The new standard, DO-178C allows for computer models to be used in verification. So we may start seeing automated prioritization and management in the generation of planes after the A380/787.

It's actually a similar situation (obviously on a smaller scale) for PC users sometimes. For example gaming hardware manufacturers have cottoned onto this fact and made keyboards with a bunch of extra keys for ingame binds/macros, because gamers would rather learn what keys do what than have any delay in their actions.

Reply on news.ycombinator.com