Adaptive UI

1
Adaptive UI

• Scott Hudson

2

• So just what are we adapting to, anyway?

3

• So just what are we adapting to, anyway?
• At least two good answers
• Devices / platforms
• Human situation

4
Adapting to devices / platforms

• Model-based UI discussion has considered some of
  this topic already (and Mark may next), but
• I think if we are going to go after this again
• (This was a staple topic in 80s tools work)
• we need to consider whether there are radically
  different ways to go about it

5
Adapting to devices / platforms

• Radically different ways to go about it
• Challenge what might those be?
• 1st suggestion numerical optimization
  approaches
• Loose collections of goodness criteria
• Range of generators
• Huge amounts of CPU time to search through space
• Run overnight or for a week on your desktop
  machine
• Because that will be 5 sec on your cell phone
  soon
• Possibly human in the loop optimization
  techniques

6
Adapting to human situation

• (For me the more interesting question)
• Current interfaces are ridiculously blind to what
  the user is up to in any larger sense

7
Adapting to human situation

• (For me the more interesting question)
• Current interfaces are ridiculously blind to what
  the user is up to in any larger sense
• Challenge develop and incorporate models of
  human situation

8
Framework I like to use

• Sense, model, and act appropriately
• Sense
• Gather the raw data (sensors)
• Model
• Make sense of it with respect to situation
• Act appropriately
• Do the right thing in the interface

9
Example from my work

• Models of human interruptibility
• Sensor-driven statistical models predict now is
  a bad time to interrupt
• Can now do this better than human observers with
  simple sensors!
• Can respond by modulating information displays
  (demand the right amount of attention) and
  communications requests (the phone doesnt ring)

10
Questions

• Are there new sensors that help a lot?
• (Raskar et al. SIGGRAPH 04)

11
Questions

• What situations (and activities) might be most
  fruitful to know about?
• What situations (and activities) look tractable
  to model (via what sensors and modeling
  techniques)?

12
Questions

• What new UI techniques are needed for acting
  appropriately?
• e.g., display methods tuned to real effects on
  human attention
• Are there tools we can build now to help?

13
(No Transcript)