Greg Detre
Tuesday, May 13, 2003
�we say
many things, but it's what we do that counts� � a girlfriend said that to him
once � it was a difficult moment :)
smart floor
in his lab
I think he�s overstating the present � last time I was there, the avatar
that gave demos, and I didn�t see the podium that appears when there�s a group
of people facing a single person
steelcase sponsor � paint sensor
where are my keys � trace path you�ve walked � couldn�t it point to them
as well � they�d have a distinctive silhouette
dynamic models of
task
system
user
none of the voice recognition systems at the moment employ micro-feedback (100ms timescale)
people can�t point as fast with a rate control device (e.g. a gas pedal) as with a position control device (e.g. pointing on a map)
allowed people to try hacking their mouse transfer function
found that if you make the thing so stiff it�s painful, it works best � not because of the greater range of dynamic control, but because you go slower than your eyes can track so that you don�t lose it, and so you don�t have the golfing problem (of a big, fast movement followed by a lot of small adjustments) � the painful one can be 25% faster
he reckons you reliably produce 4 or 5 bits worth of force control (I�m not sure of this � what about the Miller 2-3 bit unidimensional resolution) � that�s the resolution of pressure that you can maintain (rather than replicate) � ah, so there aren�t 4 or 5 bits that you can produce with different meanings � I think this is with feedback, rather than replicating on its own
so don�t make a cursor require more accuracy than that
if you make the cursor a predictable speed, and that makes it 15% faster
can�t use eyes as cursors � that�s not what eyes do � guard-dog, rememberer, attention, seek novelty, social signal
does this support or detract from my idea of using eye motion as annotation???
look at streaming headline for 1/3 second, that�s long enough to signal interest (3 times faster than a mouse)
Shieber: streaming author/subject of emails as they arrive
found that if you look at where two vertices meet, that�s a point of interest
list of sponsors of the Media Lab � the ones you find interesting start to clump together
Cheese � mouse motion
can find things you�re interested in but didn�t select � 2nd choice in list traversal
can tell whether you�ve been to that page before
presumably also what type of mouse/IO-device you�re using???
could tell between business traveller + tourist on an expedia page
a grad student at Stanford - typing monitor � showed that he was using the wrong shift keys and sped up his wpm 15% immediately
vadim gerasimov � invented tetris � designed the heartrate game where you have to pass the ball when your heartrate has gone down, having raised it before you receive the ball
one guy couldn�t go from 130-60 almost instantly � learnt that in russia to get out of the military
Driftcatcher � organised a bunch of parties with people getting to know each other � 70% confidence, only 500 messages � two groups, one with the font/labelling of email, one without, pretending to be a person�s secretary and categorising noisy data of messages � the ones with the system, even though it didn�t work that well, did much better
carcoach
vibrators on steering wheel, pedals � when to give feedback � during mistakes, soon afterwards or when you make them in a simpler situation?
his original adaptive help system made it into os2
criticism + affirmative knobs � perhaps you learn more from failure/criticism, but you change your behaviour more from affirmative behaviour
people do change their behaviour while they�re in the car � need for longitudinal study to see whether the novelty wears off, whether they just turn all the feedback off
spoon with sensors � how hard the surface, temperature, ph, salt etc.
good UI: �taking the tool out of the task�