Here’s the question: Does automatically generated interfaces — based on user skill, dexterity, and impairments — have a place in Firefox?

The University of Washington recently released a program called Supple. Essentially, it puts you through a small digital obstacle course — move your cursor here, click there — and then calculates all of your motor-skill constants: what’s the difference between your single and double click, how quickly can you target a button on screen, what’s your reaction time, etc. Once it has all of the data, it generates an interface to match your particular set of skills.

Although this loosely falls into the adaptive interface category, it doesn’t have the normal problems associated with adaptive interfaces because it only adapts once.

After reading the paper, I don’t think Supple has the ability to revolutionize interfaces per se (like the Guardian seems to think). It’s an optimization of a given set of widgets and small-scale design patterns. If you start with poor assumptions and widgets, you’ll end up with a well-optimized poor interface. And you’ll never get out-of-the-box thinking that gets you passed the point of diminishing returns on iterative improvements.

It does, though, have the ability to help us designers in making our interfaces evolutionarily better. In particular, we constantly need to base decisions on a mythical “average user”. SUPPLE can potentially let the interface mold to the actually user’s abilities, instead of the user having to become more stereotypical.

It’s ironic (and predicable) that the interfaces SUPPLE comes up with for dexterity/visually impaired people are just better interfaces than the controls. The optimized interfaces almost always display more information in a way that requires less clicking than the original interfaces. No wonder they perform better! It’s just a direct application of Fitts’s law and GOMs analysis.

For Firefox, I can see two immediate applications. The first is to optimize the current interface to best fit the user: Larger buttons in the chrome for folks with poorer motor skills (or disabilities), different preference layouts, etc. This is the kind of stuff computers are great at — constrained(ish) problems in which they can optimizing a large number of inter-dependent variables. And do it on a per-person basis.

The second is to automatically optimize the websites themselves to be better turned for the user. I have a feeling that this has been tried before, though, to marginal effect. However, SUPPLE seems to be smarter than previous attempts.

Plus, with Firefox, you never need to give the user a proficiency test. By instrumenting Firefox to do the same thing, after a day of use we’d have a wealth of information to base optimizations on. Stay tuned for more on this with Auto Pilot.

One interesting thing to call out: The interfaces for SUPPLE are defined by schematic intent, not by layout. The computer translates a user-flow markup into an actual interface. We’ll probably see a lot more of this as we need to design web sites for truly divergent screen-sizes (computer, mobile, wall screens).

RT @azaaza Personalizing Firefox’s UI automatically. Just for you. | Follow @azaaza on Twitter | All blog posts

No related posts.