Human Cognitive Abilities

1
Human Cognitive Abilities

• Daniel Glaser (Doctoral Student)
• SIMS 213, UI Design Development
• February 11, 1999

2
Outline

• What is Cognition?
• Topics of Inquiry and possible relations to
  Human-Computer Interaction
• Engineering Cognitive Modeling
• User-Centered Cognitive Modeling

3
What is Cognition?

• Cognition is how a person thinks
• It compliments perception- how a person learns
  about their environment
• Cognition is biologically and socially influenced

4
Cognition and Perception are parts of one
interrelated system

• What you know about the world influences how you
  see it and what you see influences what you know
• e.g. Our visual system will not develop properly
  in an isolated environment (experiments with
  cats)
• e.g. We ordinarily cannot see our blind spot
  (filled in by our cognition)

5
Methods of Inquiry into Cognition

• Since no one expert can say with certainty what
  other people think, cognitive science uses an
  interdisciplinary approach
• Linguistics, computer science, psychology,
  neurology, and anthropology are some fields that
  contribute to cognitive science

6
Some Topics of Inquiry

• Color (Human Visual System)
• Problem Solving
• Categories
• Spatial Cognition

Often, when describing human abilities,
scientists describe them in terms of the
computer. The mind is to brain as software is to
hardware is an implicit metaphor in descriptions
like Dix. 50 years ago the brain was described
in terms of a telephone.
7
Color (Human Visual System)

• Neurological- described in last lecture on
  perception

optic chiasma
LGN
cortex

• The further you are from the eye, the more coarse
  information is extracted from your retinal image
• Note that more information is passed from cortex
  to the LGN

8
The Visual System to Scale
9
Linguistic Inquiry into colors

• Provides further evidence that the biological
  constraints on color influence our perception

White Red Yellow Black Green Blue
White Red Yellow Black Cool
White Warm Black Cool
White Warm Dark-Cool
Light-warm
Dark-cool
3
4
5
6
2
color names
Languages with
10
Focal Colors

• People across cultures consistently choose focal
  colors (ie basic colors) as prototypes.
• Shows that certain colors are salient in peoples
  mind

Which slices look like they could be composed of
two or more colors?
11
Problem Solving (Newell and Simon 1972, also Dix
1998)

• Problem space (all possible states in the
  problem-- e.g. chess positions)
• Transition Rules (how to get from one state to
  the next-- e.g. bishop takes queen)
• Heuristics (strategies that make your problem
  solving more efficient than random-- e.g. dont
  attack queen with your king)
• Starting State (how you set the chessboard up)
• Goal State (checkmate or stalemate)

Good for well defined problems (e.g. games), but
people dont always encounter these types of
problem problems in every-day daily life.
12
Problem Framing (Rittle 1973)

• There are no definitive formulations of a problem
  (e.g. Poverty)
• Problems do not have a stopping rule-- except for
  external considerations like time and money (e.g.
  a term paper)
• Solutions are not true or false, but good or bad
  (e.g. web interfaces)

13
Categories/Properties (Dix)

• Semantic Nets

barks
dog
size medium
film character
sheepdog
collie
lassie
works sheep
14
Categories/Properties (Rosch)

• All members are not equal-- a penguin and robin
  arent equally birds in most peoples minds
• Semantic structure can be vastly different
  depending on culture (e.g. members of the food
  category)

15
Lets try a quick experiment

• People in the first and second rows, please line
  3 things up on the table in front of you (you can
  use pens, notebooks, disks)

16
Experiment, part 2

• Practice 2x picking the objects up, mixing them,
  and putting them back down in order on the table

17
Experiment, part 3

• Turn around, and put objects down in order on the
  desk behind you

18
Results of Spatial Cognition Studies in the 1990s

• Pederson, Danziger, Wilkins, Levinson, Kita, and
  Senft

Arandic, Tzeltal, Longgu Absolute- beige object
always stays at one geo-cardinal point
Dutch, Japanese, (American) Relative-- beige
object always next to right hand.
19
Implications for UI Design?

• Orientation is culturally/socially influenced

Right East (e.g. maps)
Left West (e.g. maps)
relative scale
East Hills Left Hand
West Water (Bay) Right Hand
absolute scale
20
Intrinsic Viewers too

• Participants (Kilivila, Mopan) describe tree on
  the left side of them since they place themselves
  as the character

we will return to this case later
21
Engineering Models of Cognition

• Predict User Response
• time to perform
• time to learn
• number and type of errors
• time to recover from errors
• Used after a task analysis, but before user
  testing to save costs

22
Model Human Processor (MHP)
Card, Moran, Newell (1983)
- perceptual processor - outputs into audio
storage - outputs into visual storage - cognitive
processor - outputs into working mem. has access
to - working memory - long term memory - motor
processor

• Clear separation of stages-- unlike how I
  defined perception cognition earlier

23
What is Missing from the MHP?

• Haptic memory
• for touch
• Moving from sensory memory to WM
• attention filters stimuli passes to WM
• Moving from WM to LTM
• rehearsel

Adapted from James Landay
24
MHP Basics

• Based on empirical data (word processing in the
  70s.)
• Sometimes serial, sometimes parallel
• serial in action parallel in recognition
• pressing key in response to light
• driving, reading signs, hearing at once
• Parameters
• processors have cycle time (T) 100-200 ms
• memories have capacity, decay time, type

Adapted from James Landay
25
Memory

• Working memory (short term)
• small capacity (7 2 chunks)
• 5106422343 vs. [phone]
• DECIBMGMC vs. DEC IBM GMC
• rapid access ( 70ms) decay (200 ms)
• pass to LTM after a few seconds
• Long-term memory
• episodic semantic
• huge (if not unlimited) w/ little decay
• slower access time (100 ms)

Adapted from James Landay
26
One Principle of Operation

• Fitts Law
• moving hand is a series of microcorrections
• correction takes Tp Tc Tm 240 msec
• time Tpos to move the hand to target size S which
  is distance D away is given by
• Tpos a b log2 (D/S 1
• summary
• time to move the hand depends only on the
  relative precision required

Adapted from James Landay
27
Fitts Law Example

• Which will be faster on average?
• pie menu (bigger targets less distance)

Adapted from James Landay
28
Another Principle of Operation

• Power Law of Practice
• task time on the nth trial follows a power law
• Tn T1 n-a, where a .4
• i.e., you get faster the more times you do it!
• applies to skilled behavior (sensory motor)
• does not apply to knowledge acquisition or quality

Adapted from James Landay
29
Elaboration

• Relate new material to already learned material
• Organize displays so that information can be
  chunked
• Recognition over recall-- e.g. icons, labels,
  menu names, hints, etc.

Adapted from James Landay
30
Library of Human Performance?

• Is it possible to generate a useful library of
  human performance?

Etc.

• Or will new interface design make old research
  obsolete?

31
Epistemology of Engineering Models

• Perceptual issues such as colors/font sizes
• Also, more subtle aspects of human abilities such
  as spatial cognition.
• For example, lets say you are reading a story
  about an elf on the computer. During the story
  the program refers to the item on your right...

32
Epistemology

• Is it the tree to the left of the elf, or the
  left of the viewer?

• This depends on whether or not the person makes
  an intrinsic relationship with the elf

33
Epistemology

• Inanimate object, but can one be intrinsic to it?

34
Epistemology

• Is there a left and right to a bear? If so, in
  this image are the trees in front/behind or to
  the left/right of it?

- It is unlikely that we can plug these questions
into an engineering model and get an accurate
response
35
Summery of Engineering Models

• Although engineering models categorize human
  abilities in terms of machine performance, they
  can be useful for predicting performance in
  limited domains
• Engineering models need constant revision to
  account for new user-interface technologies and
  better understanding of human cognition
• Until we have made exact replicas of people, no
  engineering model is ever complete (but that is
  not the goal of many designers)

36
User-Centered Cognitive Modeling

• Most texts on cognition usually stop with
  engineering models of cognition, but all types of
  usability testing measure consciousness at the
  individual level
• error rates/etc. Are behavioral aspects of
  inquiry
• Studying Users Mental Models is one account

37
Mental Models

• People create mental models for all aspects of
  their world.
• Unscientific, possibly contradictory, and
  unstable
• Ideally, the system designers mental model is
  compatible with how the software works and the
  user has a similar model

38
Interface Hall of Shame or Fame?

• For setting cache size in MS Internet Explorer
  (since changed)
• Slider from 1 to 100

Slide From James Landay
39
Interface Hall of Shame or Fame?

• What if you have a big disk? (e.g., 4GB
• forced to have at least a 40MB cache
• takes away control from the user
• What if they dont know their disk size?

Slide From James Landay
40
A building data structure
File system metaphor
building metaphor
41
Equivalent Information
But one diagram is more representative to the
underlying data structure
42
Example Putting It Together(Healey 98)
Height level of cultivation Greyscale
vegetation type Density ground type
43
Taking Advantage of the Viewers Cognitive System
A better base map would have state lines drawn
(since I can associate vegetation with states),
and terrain (ground type)
If your viewer lived in America-- seen drawings,
eaten these foods (visually identify them), know
something about geography (states), this may be
more clear
44
Is there a way of combining these methods for a
more understandable display?
easy associations - data guessing - (does not
show full data set)
accurate data display - difficult to associate
data with meaning