Adaptive User Interfaces

Years ago, back in the days of Windows 95 and Mac OS 8 I had an interesting idea for providing a better user experience by utilising adaptive user interfaces.

Ten years later, after a rather surprising encounter with a particularly badly designed interface element in Vista, I am again reminded of the idea and thought I would mention it, because it appears that despite being discussed in academia it is yet to be explored or adopted in the commercial space, and the potential use has increased dramatically.

You see, both Windows and Mac OS have control panel settings to configure things such as what the double-click rate should be, what the keyboard repeat delay is, and how fast the mouse cursor moves and accelerates. Other items include the width of window borders that users grab, and the delay before side menus fly out.

I found it interesting that advanced users would immediately jump to the control panel when using a new machine (or a shared machine for the first time) and configure these to the upper limits, wanting snappy, calisthenic interactions to make their lives easier and more productive. I also realised that novices or first time computer users needed large windows of time between the two clicks in a double, and that their mouse accuracy would demand less sensitivity (and I was always annoyed at Apple’s lack of support for mouse acceleration, but that’s another story), hence the existence of these controls in their default settings.

The thing is, users evolve and get better at using input devices, and it struck me as somewhat unintelligent to require users to have configure these parameters to suit their interaction ability either when they start using a computer, or as they mature as user. Surely a smart OS would set the limits to the lowest by default, and analyse the behaviour of the user and adjust the settings to suit? This way, when I buy a new computer, I simply turn it on, and it becomes aware within minutes that someone with a fast double-click rate, and a minimum of overrun when locating icons I am a user wanting tighter limits and a more sensitive mouse. Likewise, novices get an initially forgiving interface that adapts and develops as they do.

I think it’s high time for this type of intelligent, adaptive user interface that acknowledges and adapts to a user’s interaction ability. However, there also exists a new dimension to this concept, that is, today’s connected nature of computing. Which nicely segways into the the Vista example in question. May I present to you, ladies and gentlemen, one of the most depressingly bad UI design decisions since Genghis Khan opted to have poisonous spikes installed on the inside of all his soldier’s helmets.

This is what Vista presents me when I try and copy a file into a folder containing a file with the same name.

Just as the web finally gets round to the notion that people don’t really read pages online, here we have an OS interface element deciding that it should look and feel like a rather verbose (and badly designed) web page.

I’m not going to get into a rant about the grossly disappointing nature of this dialogue box - how there’s only one obvious action and that’s cancel, and that it takes around 40 seconds to read the text, let alone comprehend my options. That it’s full of entirely accurate yet totally useless information that does nothing more than confuse the user. It’s UI practices like this that throws a dubious light on the whole of Vista (if they let this one get in, what else have they got so wrong?), and calls me to question how much thought, wisdom and testing actually did go into MS Office’s new ribbon UI. It also makes me wonder how, after all these years can an organisation as large as Microsoft, as full of amazingly intelligent people as it is, still get such a basic, but major, fundamental thing so, so wrong. It’s dialogue boxes like this, that make users stare blankly, design intelligentsia scoff, and Steve Jobs gleefully chuckle at fodder for another ‘switch’ ad. Oops, that sounded awfully like a rant. Sorry.

Sure, I might be overstating this a little bit, but it’s just a sad example of how little we’ve progressed in HCI. It’s the year 2008 and things like this are still being churned out. I actually don’t want to discuss what is wrong with this form, and how to fix it. Clearly someone at Microsoft thinks it’s fine, and yet I think any HCI graduate (hell even a smart design student) could roll off required best practice to fix this without to much trouble. It seems likely to the observer that this form would be the work of any overzealous developer who wants to demonstrate the fantastic functionality available in Vista. It should be the designer or usability professional’s job to take into account the user and HCI principles and refine this down to an appropriate interface and interaction model. What I want to do is discuss why didn’t this happen at MS? What are they missing? What is their excuse? #### I’m going to go out on a limb and suggest that (apart from their undying loyalty to the developer) their only excuse can be that the interface was never properly tested. I have to assume this because it flies in the face of too many classic UI principles plus it is actually a right pain in the arse to use. It’s sad that there wasn’t someone at MS who didn’t instinctively know this was not good design, but it’s sadder still that it made it out into the wild.

Now for companies like Microsoft, user testing in the lab should be obvious and a matter of course. But there is another far better option - real-world measurement and analysis of actual user behaviour. This should be relatively trivial for Microsoft, because they’ve got a rather healthy user-base. The relative success or failure of this dialogue design could be easily settled through this type of remote measurement.

It would be easy to measure and decide on the success of a design through A/B testing, or analysis of it’s use, or the user’s behaviour interacting with the dialogue. For instance we could see how many people actually click an option other than Copy and replace (a horrible label btw), how many people mistakenly click cancel and then repeat the whole process over again. We could measure how long the form takes to be read and understood by users, and compare against other form permutations.

Of course it would be so much more powerful if remote measurement was used to float and test genuinely innovative interaction models, as opposed to things are rudimentary as this type of dialogue box. But hey, you’ve got to start somewhere right?

I hope that future operating systems do a better job out of the box, but I think that really they can go so much further in this regard. I think that the next generation of intuitive user experience design should enable adaptive user interfaces that tailor to suit individual user’s skills and abilities, and secondly, they should also leverage interaction measurement and analysis from millions of monitored incidents in the wild to better understand how users are using interfaces in front of them.