Slashdot Mirror
User Interface Design for Programmers
Spolsky's light writing style makes this book an easy read, and his personal stories and anecdotes help make his thoughts on user interface stick in your mind when you're done reading. He provides programmers with a few simple guidelines to follow, such as "People Can't Read," and "People Can't Control the Mouse."
His focus on the logic of good user interfaces and his push to develop a good user model is bound to resonate and get programmers to think about making their interfaces logical from the user's perspective, rather than the perspective of the inner architecture, which the user could typically care less about.
The reminder to focus on the tasks the user is trying to accomplish rather than the long feature list that usually gets attached to product specifications should be read by product managers as well, of course. In fact, the absence of specific platform details makes the book a good read for anyone involved in software design -- with the caveat that it is not aimed at people with much design experience. This is a great starter book and makes the process understandable, friendly, and fun-sounding. (One of the things I appreciated was how much fun it sounds like Spolsky has when he's working.)
Spolsky encourages showing the in-progress software to users and watching them use it. I think one of his best points about usability testing is that if the programmers and designers cannot bother to watch the users during the testing, they're unlikely to gain much from a thick report by a testing lab. He encourages simple, quick, and casual usability testing, something even the smallest firm could afford and from which they would could draw useful improvements.
If you have much design experience, you'll find this book a bit basic, but even then the examples are worthwhile to read and remind yourself how a good idea can be poorly implemented sometimes -- usually by taking it too far! I was personally hoping for some richer comments about designing web applications, but if more people start paying attention to the basic guidelines he's covered here, web users will benefit.
In summary, the book is aimed at programmers without much design experience and Spolsky does a great job of hitting his mark. I think product managers without much design experience would benefit as well, as it provides a good basis for thinking about user interface design.
You can purchase User Interface Design for Programmers from bn.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.
they consider it a creative process rather than a logical one;
Are we supposed to assume that creative and logical are now mutually exclusive? I always thought they were complementary. I sure as heck wouldn't find computers interesting if it was all rote and mechanics.
What is music when you despise all sound?
Good UI design just takes a lot of time and a lot of listening. First, you design the interface to do what you want it to do. You try to pretend you know very little about the actual mechanics of what gets done behind the scenes to make whatever it is happen (a difficult proposition, but you should be able to get relatively close). Then, code the interface (just the framework, don't waste a whole lot of time at this point).
Then, show it to someone representative of the intended audience. If you're coding a general purpose Windows app, show it to your grandmother. See if she can figure out how to work it. Encourage conversation about it. If she can't figure it out, don't get argumentative. Find out what SHE thinks the interface is trying to do, and try to find out what about the interface makes her think that. Then, try to get a few ideas on how to improve it. She won't be able to give you any real specifics, but maybe she can give you a thread you can explore in detail on your own.
Re-design based on what you learned. Show it to her again. Repeat until she "gets it". Then, go show your new design to someone else in your target group. Make changes by what they say. If what they say contradicts what your grandmother said, do your best to reconcile the differences. Make up any gaps you can't fix with documentation targeted at the bits you can't seem to make any less confusing.
A lot of engineers fall into the trap of designing interfaces and sticking with them, even if they are deficient. They insist the users are just "too stupid" or just "don't get it" or just "aren't using it right". They fail to realize the whole idea of a good UI is to make sure users CAN'T use it wrong, and to make it as difficult as possible for the user to fail to understand.
"The customer did something wrong" is NEVER a reasonable excuse for a problem in a UI. If the customer did something wrong, it's YOUR fault for making it possible for the customer to do whatever it was he did wrong.
Way back in days of yore, when Microsoft was still working out how to do overlapping windows, there was a company called Geoworks that produced a really nice office suite for the PC.
I won't go into details about it, but one of the really cool features was that each application had a tunable user interface. For example, you could set the word processor to user level #1 (novice) and it would turn into Windows Write: most of the controls went away, and you ended up with toolbar buttons for italic, bold, underline, etc, plus justification options; you got simple menus that let you pick things like the font and size directly; you got really, really basic page layout features --- I think it let you pick your paper size, and that was it.
OTOH, turn it up to level #4 (expert) and it turned into Word. There were controls everywhere. Hierarchical editable character and paragraph styles, embedded fields, hyperlinks, a full vector drawing package including rotatable text (also with hierarchical editable styles), a full bitmap drawing package, up to four seperate customisable toolbars, ruler and frame based layout, etc, etc.
And they used the same files.
So it was perfectly possible for Precocious Teenager to log in in expert mode, put together some pretty templates, and then Grandma could log in in novice mode and type text into them with simple formatting. Mum and Dad could use levels #2 or #3, which gave you more features without the overwhelming complexity that level #4 gave you.
It was such a startlingly good idea that I am not at all surprised no-one appears to have done anything similar.
(Hmm. You might still be able to download an evaluation copy here, but I suspect it's a pig to run on a NT-based Windows. Worth a look, though, if you want to be amazed at what it's possible to do on a 2MB real-mode DOS machine.)
Users rarely complain about badly designed user interfaces. They accept that computers are nasty, evil devices that make their lives hell and prevent them from doing work as much as possible. They say nothing to you, and then they come home to their families and say "I hate computers".
An end user not complaining about a bad UI is like somone complaining that a torture device like the rack is "uncomfy". It's just accepted that the experience will suck.
There is so much general computer-phobia in the world because end-users have not yet realized that it's not the computers in general that's the cause of their problems with an application, but rather it's the individual programmers who wrote the application who are the problem.
Ergonomica Auctorita
Ergonomica Auctorita Illico!
Unfortunately, UI can also be an area that should *not* be consumer-driven.
The recent facination (last five years) with media player authors to make "pretty" interfaces that immediately grab a user's interest is a great example. The UIs are far less usable, are inconsistent, are frequiently slower and buggy...yet authors keep pumping out these damned bitmap interfaces to DVD players, movie file players, audio file players, etc.
The problem is that every time someone does something with a tiny bit of justification, everyone copies it wrong.
Bitmapped interfaces have seen two major insurgences that are still present. The first, pointed out earlier, was in media player apps. There are a number of cases, but I think the first instance I know of was WinAmp. WinAmp was trying to fill a hole that had never been filled before. It needed to remain constantly up on a user's desktop to keep title, volume, and position available. However, it needed to save space (see the minimized form) -- I can't think of a good way to provide equivalent functionality using standard widgets. Anyway, a difficult HCI call -- to deviate from the standard OS interface was made. It has definitely had drawbacks, but there's at least a good argument that it was worthwhile.
Along came a huge number of media player designers, all of whom looked at WinAmp and decided at the bitmapped interface was what made the thing successful. They started churning out all kinds of horiffic unusable media players that *did* catch the eye, and *did* get users to try them out...only to hit irritation with the interfaces. Media players pioneered spikes hanging off of windows.
The other major example is graphic plugins, dating back to Kai's Power Tools. For those not familiar with the tool, KPT is a set of Photoshop plugins. It was written by Kai Krause, an extremely talented graphics programmer. He felt that using custom bitmapped widgets was a good idea. Again, his decision was somewhat arguable, but it let him showcase some of his software's effects, and more importantly, he did a reasonable job for someone going with an inconsistent interface -- he did a few things that would have been difficult with a conventional widget set. KPT had a tremendous functionality set, and succeeded wildly, allowing the company to grow, change names, and develop and acquire other software products like mad. The company continued to produce other outstanding products, also with bitmapped interfaces (with greater and lesser degrees of justification for their nonstandard interfaces. KPT Bryce is a notable example.
Naturally, a number of other, less talented, Photoshop plugin development companies that were producing products that were not particularly price-competitive or feature competitive looked at KPT and said "Gee...KPT uses a bitmapped interface and is successful. That must be what we're missing." Over the next few years, a *flood* of inconsistent, bitmap-interfaced Photoshop plugins hit the market. These were, as a rule, less well-done than the original KPT, and were a complete pain in the ass for a set of people that mostly used Macs, and had traditionally had one of the most consistent user interfaces in the history of personal computing.
Bitmapped, custom interfaces are almost always a bad idea.
There was also an influx of CD-ROM based titles with bitmapped interfaces starting in the early CD-ROM days. Lots of low-budget titles, educational titles, etc. Macromedia Director played a major role in the proliferation of these. Again, a bitmapped interface added nothing to usability, and frequently exposed bugs. It took a few years, but eventually designers realized that users didn't *like* atrocious bitmapped interfaces, and stopped.
Today, almost all games have a menu system that uses a nonstandard, bitmapped interface. Part of this is because they often have console ports, where there *is* no standard widget system, and part of it is because there's a perception that the customer *wants* a m
May we never see th
Windows isn't any better. Sure, CTRL X/C/V are fairly standard, but anything more than that is terrible.
Want to do a "find"? Well, it's CTRL-F... usually. Unless you're in Outlook, where CTRL-F does forward, and find is (intuitively!) F4. Oh, except for the main message list, where Find doesn't have a shortcut at all, but advanced find is CTRL-SHIFT-F. And don't get me started on third party apps like Textpad (which is a great app, but uses F5 for find and F8 for find/replace).
Button location is another bugbear. OK and Cancel randomly move around dialog boxes, swapping positions with merry abandon. Always assuming they're present, of course - dialogs are sometimes closed with "Ok", sometimes with "Close", both doing the same thing (often in the same application. Sometimes there's a close box, sometimes not.
A much more consistent interface is the mac, for historical reasons. Find is always CMD-F in every major application. Closing a window? Always CMD-W. Quit an app with CMD-Q. When it comes to dialog boxes, Apple doesn't just specify the names of buttons - they tell you where they should be placed (to the pixel), how they should work, what types of icon should be shown for each type of alert and so on. Sure, apps don't need to follow the guidelines - but they pretty much all do, simply because anything that doesn't just looks "wrong" to mac users who are used to consistency.
It always bugs me when I see linux advocates pushing coders to take Windows as an example of a good interface. It's a dreadful interface (admittedly much improved recently), and despite Apple's recent minor UI setbacks in OS X, it's still by far the best designed interface available. Don't just copy the style - if you understand why the mac interface was designed that way it was, you'll be able to produce something nicer than 90% of apps on any other platform.
'Options' are good case in point. Often people want extra options to un-break some poorly chosen UI behaviour or functionality. It is beter to find out what is really causing the problem and fix that.
--
Simon