HCI Lecture 2: Human Capabilities: Input/Output Systems: Key Points

HCI Lecture 2:
Human capabilities: Input/Output systems
Hiroshi Shimodaira
Key points:
  Human have processing constraints
  Motor limitations, e.g. Fitts’ law for pointing
  Visual range for motion, shape, colour, detail and
their consequences for design decisions
  Visual attention models
  Alternative sensory channels
1
Human constraints
  Human computer interaction depends on what humans are

actually capable of observing and articulating
Task
Articulation Performance
Input
Human Computer
Observation Output Presentation
Interface
Environment
2
Human constraints
  What do we know about human capabilities that could or

should constrain interface design?
  Limits on perceptual capability – e.g. contrast, resolution

  Limits on motor capability – e.g. reach, speed, precision
  Limits on attention capacity
  Limits on memory
  Rates of learning and forgetting
  Causes of error
  Mental models & biases
  Individual differences (the average size fits few people)
  Variable state (e.g. stress, fatigue)
  Special needs & age …
3
Human constraints
The MHP
  Model Human Processor

(MHP)
  One way to subdivide
the main constraints
  Perceptual, Motor and
Cognitive sub-systems
characterised by:
–  Storage capacity U
–  Decay time D
–  Processor cycle time T
  Wewill focus today on

the perceptual and
motor processes
4
Motor constraints
  Example: Fitts’ law (1954)
T=a+b log2(D/W+1)
T=time, D=distance, W= target width
a, b are constants that depend on the pointing device, the user, the
environment etc.
5
Motor constraints
T=a+b log2(D/W+1)
environment etc.
  Justification?
  By “analogy” to Shannon information
capacity = bandwidthxlog2((signal+noise)/noise)
  If move fraction 1-r to target each timestep, then reach
target when rnD = W/2; so n is proportional to log22D/W
  Empirically find good fit with log2(D/W + 0.5)
6
Motor constraints
T=a+b log2(D/W+1)
environment etc.
  Application?
  Time will increase with distance – can we keep everything close?
  Time will decrease with width – can we make width infinite?
See quiz regarding HCI applications of Fitts’ law here: http://

www.asktog.com/columns/022DesignedToGiveFitts.html
7
Perception
  What can we see?

detail
colour
shape
motion
8
Perception
  Some consequences of what we can see:
  Motion – will be visible (and distracting) anywhere in visual field
  Colour – main advantage is “pop-out”:
But many disadvantages:
  Shape important in text recognition: SO ALL CAPS BAD
  Limits on resolution – recommend minimum font size; ideally individual can adjust
  High resolution only in tiny area of fixation
9
Eye tracking
  Fixation pattern is a good indicator of attention

  Where do people look, how often, for how long, in what order?
  Recent technology is making this a standard tool for HCI
Babcock &
Pelz 2004
  Also used as input device.
10
Perception
  Importance of eye movements

  Must shift the tiny high
resolution area around
constantly
  Movements called saccades
occur > 2 per second all day
long
  How does visual system decide
where to move next?
  Models of attention
  e.g. Itti et.al. 1998
11
Attention
  Simple statistical model of saliency

Rosenholtz et al (2005)
  Provides definition of ‘clutter’: size of
local covariance ellipsoid
  To measure:
  Compute local feature covariance at
multiple scales
  Take maximum across scales
  Average for different features
  Pool over space
  Produces good correlation with human
estimates of clutter
  Can also use to determine what
feature added where would best draw
attention
12
Attention
  So what went wrong here?

  Task: find current population of U.S.
  86% of users failed…

http://www.useit.com/alertbox/fancy-formatting.html
13
Perceptual constraints
Bottom up visual processing sets some constraints on optimal

layouts, but must also consider top down issues:
  Cultural and learned factors – familiarity
  Underlying domain knowledge of user
  Need to reflect logical structure, e.g., placement and grouping
according to function, sequence, frequency of use
  Dependence on task to be carried out, e.g. getting an overview
vs. seeking specific information
  Note that layout and visualisation are already widely explored

fields, with conclusions that carry over to HCI
14
Alternative sensory channels
  Different sensors provide parallel channel capacity
Sound:
  Not so easy to localise but can detect from any direction
  Grabs attention – warning mechanisms
  Good signal of causal relation – use as confirmatory feedback
  Monitoring state, ‘background information’

  Disk, printer noise etc.
  Example of user improvisation in use of ‘data’
  Interface sound design is typically arbitrary and synthetic
Touch and haptics:

  Exploit our natural ability to ‘handle’ objects
15
References
Fitts law: for a detailed account see MacKenzie, I.S. (1992) Fitt’s law as a
reseach and design tool in human-computer interaction. Human
Computer Interaction, 7, 91-139.
Itti model of visual attention: see ilab.usc.edu/bu for details of the model,
images, movies, an interactive demo and source code.
Jay, C et al. (2007) How people use presentation to search for a link:
Expanding the understanding of accessibility on the web
www.cs.man.ac.uk/~jayc/papers/web_presentation_new.pdf
Rosenholtz, R. et al. (2005) Feature congestion: a measure of display

clutter. SIGCHI 2005, 761-770
  See also:
Dix et. al. sections 1.2, 3.2, 3.4, 12.5
16

HCI Lecture 2: Human Capabilities: Input/Output Systems: Key Points

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

HCI Lecture 2: Human Capabilities: Input/Output Systems: Key Points

Uploaded by

Copyright:

Available Formats

HCI Lecture 2:

Human capabilities: Input/Output systems

 Human computer interaction depends on what humans are

 What do we know about human capabilities that could or

 Limits on perceptual capability – e.g. contrast, resolution

 Model Human Processor

 Wewill focus today on

 Example: Fitts’ law (1954)

 Example: Fitts’ law (1954)

 Example: Fitts’ law (1954)

See quiz regarding HCI applications of Fitts’ law here: http://

 What can we see?

 Some consequences of what we can see:

 Motion – will be visible (and distracting) anywhere in visual field

 Colour – main advantage is “pop-out”:

But many disadvantages:

 Shape important in text recognition: SO ALL CAPS BAD

 High resolution only in tiny area of fixation

 Fixation pattern is a good indicator of attention

 Also used as input device.

 Importance of eye movements

 Simple statistical model of saliency

 So what went wrong here?

 86% of users failed…

Bottom up visual processing sets some constraints on optimal

 Note that layout and visualisation are already widely explored

 Different sensors provide parallel channel capacity

 Grabs attention – warning mechanisms

 Good signal of causal relation – use as confirmatory feedback

 Monitoring state, ‘background information’

Touch and haptics:

Rosenholtz, R. et al. (2005) Feature congestion: a measure of display

You might also like

  Human computer interaction depends on what humans are

  What do we know about human capabilities that could or

  Limits on perceptual capability – e.g. contrast, resolution

  Model Human Processor

  Wewill focus today on

  Example: Fitts’ law (1954)

  Example: Fitts’ law (1954)

  Example: Fitts’ law (1954)

  What can we see?

  Some consequences of what we can see:

  Motion – will be visible (and distracting) anywhere in visual field

  Colour – main advantage is “pop-out”:

  Shape important in text recognition: SO ALL CAPS BAD

  High resolution only in tiny area of fixation

  Fixation pattern is a good indicator of attention

  Also used as input device.

  Importance of eye movements

  Simple statistical model of saliency

  So what went wrong here?

  86% of users failed…

  Note that layout and visualisation are already widely explored

  Different sensors provide parallel channel capacity

  Grabs attention – warning mechanisms

  Good signal of causal relation – use as confirmatory feedback

  Monitoring state, ‘background information’