Slashdot Mirror
Jef Raskin Talks Skins
gwernol writes "Jef Raskin, one of the original Mac design team and a distinguished figure in the world of user interface design has given an interesting
interview over on OS Opinion. He talks about the tradeoff between interface consistency and customizability, and particularly the impact of skinnable applications on usability. Interesting reading, including some harsh words for "guru UI designers" like Steve Jobs..."
Look up Project Ernestine.
;-)
A group of researchers applied GOMS (which is a form of task analysis) to a new workstation designed for use by telephone company toll and assistance operators.
The new workstation should have been faster, according to those that designed it, because it:
Ran on a faster network
Had a 'better' GUI
The keyboard was remapped and some of the functions moved to allegedly speed up the operations.
and so on.
However, task analysis (and real-world testing) showed that the new system was in fact slower. There are technical reasons for this, eg. that although the network ran faster, the original system had redrawn line by line and therefore the operators had not needed to wait for the screen to completely draw... but partly, it was that some of the alterations that they had made were about as useful as feet on a fish.
A 'better gui'? What does that have to do with telecom operation? And many of the changes they'd made to the keyboard had taken operations that they would originally have done in 'slack time' and placed them in the critical path so that the operators actually had to type faster...
All this doesn't prove that command lines are necessarily faster.
What it does show is that many of the assumptions made by those who try to design 'better interfaces' are wrong - eg. the GUI - and that if you want to design an efficient user interface, you absolutely have to do it to suit a particular user or class of users. For a different user - say, an untrained beginner to the job - the Project Ernestine interface might have been far easier to use.... and therefore initially more efficient...
*sigh*
UI research is a hard problem. In my personal opinion, the current state of research is seriously broken in a number of ways - how do you measure the usability of a program? Why, you measure its efficiency! Um... but maybe I'm not looking to get my image drawn 0.56 seconds faster, but to get some artistic inspiration going? Well then, the 'state of the art' choice is probably heuristic evaluation, which is virtually empirical and just about has a sort of scientific basis. And it certainly won't tell you just how happy a user is. At which point it all comes down to using questionnaires. And at that point, you might as well kiss all this scientific theory stuff goodbye completely...
We tend to think of software as a tool, and computers as the beepy box on which those tools rely. I feel that this devalues the computer. User interfaces, in my uninformed opinion, tend to hide the computer (the freedom, as with the command line, to create more tools, fluidly) behind buttons, toolbars, and predefined courses of action. But what do I know?