First Session: Short Talks on Input by Hand, Eye, and Brain

The Two-Handed Desktop Interface: Are We There Yet?

Interesting idea, but what this boiled down to was a lefthanded person trying to make a fifteen-minute presentation out of the fact that he could use a mouse and keyboard maybe a little more efficiently than a right-handed person. No data, no experiment. Makes you wonder about the review panel. The point was still valid, though:

• We use our non-dominant hand to lead action, to set the frame of reference, and for coarser actions
• We use our dominant hand to follow, to work within the frame of reference, for fine action
• The executive keys on the keyboard (enter, delete, insert, backspace, etc) are on the right side
• so, if you're left-handed, you can mouse with the left hand and still hit useful keys with the right hand. If you're right-handed, you can't do this as well.

all in all, this struck me as a very Canadian presentation.

Natural Hand Writing in Unstable 3D space ...

A Japanese-English language barrier made this a bit tough. The bottom line was that it's hard to write or draw in midair. Using a prop like a wooden block makes it much easier. No real experiment or data, but at least this one was based on some actual mucking around in the lab.

Real-time 3D interaction with ActiveCube

A Japanese researcher has a set of cubes that snap together. Cubes include a camera, ultrasonic distance detector, light, buzzer, gyroscope, and so on. When you snap them together, a computer model of the block assemply updates automatically. I think we were all (metaphorically) looking at each other dubiously - what's the point? The point was make very effectively, however, when the demonstrator showed a video: First, he connects a light to the sensor. When he moves his hand towards the sensor, it gets a bit brighter. And in reverse. Then he hooks up a buzzer. Now, as he moves his hand back and forth, a piercing and unpleasant whine rises and falls. (In the background, a computer screen has faithfully recorded the presense of the buzzer module.) Then, he adds a vibrator module. Now a blinking, whining block contraption buzzs around on the laptop keyboard and table surface in the demo video, threatening to pitch over the side of the table, as the audience in the room finally loses it and we all burst out laughing. What we have here is a new form of chindogu, the 'un-useless invention'. http://info.pitt.edu/~ctnst3/chindogu.html. In response to a question, the speaker tried to list some possible practical applications but nobody was buying his ideas.

Solving Multi-target Haptic Problems in Menu Interaction

ok, this one was non-bogus. A scottish team worked on a force-feedback 3-d pen (a sort of robot arm with a pen at the end, called PHANTOM http://www.sensable.com/haptics/products/phantom.html) to see if haptics (the sense of touch) would help people navigate menus. Previous research has shown that Haptics does help the single-select case.

That is, if there is one button on the screen, and the subject has to move the mouse across the screen and click on the button, it helps if the button actually sucks the pointer towards it (via physical pressure on the pointing mechanism). However, this doesn't work for multi-select cases, because you have to drag the pointer across a bunch of sticky 'wrong choices.' One fix is predictive: guess which button the user wants and make the others non-sticky. But that takes a lot of horsepower and still can't be done well (and if you can accurately guess what the user wants to do next, why the hell are you making them drag a cursor around a sticky screen?).

So. They tested the Windows Start menu, requiring users to navigate a few cascading menus to pick the right item. In the control, there was no feedback. In the 'Haptic' case, each menu choice was sticky as described above. In the 'Adapted' case, the force feedback was orthogonal to the direction of movement and proportional to speed. That is, if you start moving horizontally, it gets hard to move vertically.

And the result, with 18 subjects, is that the 'Adapted' case was fastest and generally more accurate, compared to both the visual-only control case and the 'Haptic' simle sticky case. woo-hoo!

Vision-Based Face Tracking System for blah blah

Current systems of vision tracking are intrusive (you have to put something on your face), and take a while to calibrate. (an audience member later argued that these points are less valid now than they were a few years ago). This team devised a system for determining the face orientation (not the pupils) based on the corners of the eyes and mouth, which are easy to find and which have a constant geometrical relationship (as long as you are expressionless). They experimented and found that this system made it easier to control scrolling and zooming of a map application, but wasn't great for task switching. The demo video was actually a bit startling: the user leans forward and the map zooms in.

Nudge and Shove: Frequency Thresholding for Navigation in Direct Brain-Computer Interfaces

On the minus side, this presentation only presented something resembling experimental results for a few minutes at the end. On the plus side, out of the entire three-day conference, it was far and away the most intrinsically interesting subject, and the one most fundamentally appropriate to a conference about Computer-Human interface.

Melody Moore, a Ph.D. at Georgia Tech, claimed "We're the only team in the world approved by the FDA to implant sensors [of this type] in human brains." They have placed electrodes) into the brains of three subjects, all suffering from 'locked-in syndrome.' That is, they are a few of the 500,000 people worldwide who are essentially completely paralyzed, cannot live without continuous machine assistance, but are cognatively fully capable.

The best subject of the three has had electrodes in his brain for three years. One controls the X and the other the Y for a cursor. Glass capsules in thecortex communicate via FM with chips implanted in the scalp (outside the skull) and then to computers. He can move a cursor around the screen and type at 3 letters per minutes on a virtual keyboard (previously he used blinking to select letters - she didn't say what his speed was for that). Every time they experiment they ask him what he's thinking about while moving the cursor. One day, he said "nothing." This means that he controls the mouse with the same mental method that we use to move our own limbs. He has a "cursor cortex."

They are fully funded and approved for six more subjects and are looking for volunteers.

Second Session

I ended up skipping around, catching a half-hour each of three events, none of which really grabbed me. A workshop on Promoting Usability Practices within the Design Process was basic and boring. One useful bit: a list of other sources of usability defects

• newsgroups
• remote testing
• product instrumentation
• user group meetings
• support/service
• training
• field reps
• documentation writers
• marketing

I caught a bit of something about animation, a prototype for combining Back, History, and Bookmark in one panel (good point: many users don't really understand how Back works. Browser back works on a stack model, not an absolute model, so it doesn't track branches, and this trips up users.), a vaguely interesting demonstration of a smartmoney.com feature where you draw a graph to query time-series data for best matches; and sat in the back for the end of The Impact of Mobile Technologies on Everyday Life (SMS hasn't taken off in this country because our teenagers already all have phones in their own rooms with free local calling, btw, and because the American pager network has been up and reliable for a long time). oh yeah, and "Text input is the fundamental problem" for mobile computing. I was pretty tired by this point.

Third Session

Information-Rich User Interfaces

this seemed promising but turned out to be fairly low-bandwidth (or I was just really tired) and I bailed partway through. Don't have "duh" help that just repeats text already on the screen. On the other hand, there's nothing so simple and easy that there's no meaningful help. For example, "First Name" - what if I have an umlaut in my first name? Examples are often the most useful, information-rich help. You can have data-sensitive help, such as "Your hard-drive is 50% full, ...." And also use-sensitive help, "You just hit six menu commands in the last 30 seconds. Perhaps you need help."

Information Scent

I missed most of this, but the bit I caught seemed promising, albeit strongly biased towards an academic perspective. Information scent is the amount by which a given page makes a user feel that they are getting closer to or further from what they're looking for. The idea is that you can somehow measure information scent (by hand, in these examples) and see how users follow it. Turns out (with the pretty nifty data and charts) that when information scent is poor, users jump to a lot of pages and go in circles. When it's strong, they zoom right towards what they want.

When scent falls below a threshold, users change something, either switching modes (such as to a search) or browsing from a different starting point.

Closing Speech

• Designing for universal access doesn't just mean designing for the handicapped. It includes compromised situations (noisy, like factory floor; distracting, like while driving; with one hand full; etc)
• Frequently it's just not possible to serve everybody with one design. But don't give up on that before you've tried.
• Many inventions (keyboard, ramps on curbs, forgot the others) were originally built for handicapped.
• General population is getting older and older; universal access will only get more important.

More General Comments

While the conference body seemed to have gender parity (and participants from six continents), the various committees that went on stage to be applauded were still pretty unbalanced. And out of 3000 attendees, I counted less than ten black people.

It's pretty rare for usability to have true buy-in from management. I didn't see any real mention of continuous usability testing.