Title: The Roles of Knowledge and Display Characteristics in Interpretation of Weather Maps
1The Roles of Knowledge and Display
Characteristics in Interpretation of Weather Maps
- Mary Hegarty
- University of California, Santa Barbara
- Thanks To ONR,
- Naomi Shimozawa, Matt Canham, Dustin Calvillo, N.
Hari Narayanan, Joel Michaelsen, Ted Tsui
2Model of Graph(ics) Comprehension (Pinker, 1990)
Conceptual Message
Early Visual Processes
Match
Encoding
Graph
Visual array
Visual Description
Inference
Graph Schema
Conceptual Questions
3(No Transcript)
4(Implicit) Assumptions about Knowledge and
Graphics Comprehension
- More knowledge leads to better comprehension,
particularly inferences - If you can encode information from a display and
have the relevant knowledge to make an inference
from that information, the inference will be made - Information encoded and inferred is propositional
5Other Possibilities
- Knowledge and correct interpretation may not be
sufficient conditions for an inference to be
made. - May not activate correct knowledge
- Propositional knowledge vs. visual-spatial
representation - Naïve theories may interfere with correct
principles - Characteristics of the graphics (e.g. complexity)
may affect inference process
6Meteorology as a Domain
- Data/displays can be arbitrarily complex
- Visual displays are typically used to make
inferences (predict weather) - People vary in their amound and source of their
knowledge - Newspaper/weather channel
- Direct experience of weather
- Formal instruction
- Forecasting experience of experts
7Introduction to Our Weather Prediction Task
The weather at X is getting warmer. True or false?
X
8How Can you Tell?
L
H
9How Can you Tell?
L
H
Pressure Gradient Force (PGF) Air moves from
high to low pressure.
10How Can you Tell?
L
H
(Northern Hemisphere)
Pressure Gradient Force (PGF) Air moves from
high to low pressure. Coriolis Effect Air
moves counterclockwise around low, clockwise
around high pressure.
11How Can you Tell?
- Step 1 Infer wind direction from pressure
Step 2 Infer temperature change from wind
direction
12- Air will move counterclockwise inward -
Weather at X will get colder
13Experiment 1
- Participants
- 16 novices (geography) students
- 15 naïve (psychology) students
- Tasks
- Meteorology Questionnaire
- Weather Prediction
- Display
- Separate temperature and pressure maps
- Combined temperature and pressure map
- Temperature, pressure and cloud cover map
- Accuracy, RT, verbal protocols
14Separate Maps
Will the weather get warmer or colder at X?
15Integrated Map
Will the weather get warmer or colder at X?
16Complex Maps
Will the weather get warmer or colder at X?
17Questionnaire Data
18Weather Prediction Accuracy(Proportion Correct)
Chance
19Weather Prediction Reaction Time (seconds)
20Heuristics Identified in Protocols
(1) Pressure differential (Air moves
from high pressure to low pressure.) (2) Coriolis
effect (Air moves counterclockwise
around low pressure,
clockwise around high pressure.) (3)
Temperature-Pressure Associations (Low
pressure is cold. High pressure is warm.) (4)
West-East flow of air (Air moves from
west to east.) Other Unable to classify.
21 Trials Principle mentioned
22 Trials Principle mentioned
23 Trials Principle mentioned
Students using heuristic consistently
24Correct answer consistent with all 4
principles 83 accurate
25Correct answer inconsistent with 3 principles 23
accuracy
26Summary (Experiment 1)
- Novices had more domain knowledge but were not
better at weather prediction - Weather prediction performance close to chance
- Performance faster and more accurate for
integrated maps compared to separate and complex
maps - Novices more likely to base answers on principles
(not always correct principles)
27Experiment 2a Experts
- Is there a right answer?
- 3 Meteorology Experts
- NRL Monterey
- Ph.D. in Meteorology
- 15 years experience
- 2/3 had experience as forecasters
- 10 weather maps
28The area is getting colder.
- Draw arrows indicating how the wind will blow
around the target area. - Is the statement True or False? (Please circle
true or false.) - Explain why you think so. Please write your
explanations as clearly as possible.
29- Arrows counterclockwise inward - Agreed on
correct answer
30Experiment 2b Naïve Individuals
- Taught Naïve individuals the Pressure Gradient
and Coriolis principles - 3 groups of 16 participants
- Control
- Learned Principles
- Learned Principles and worked-out examples
- Measured
- Pretest posttest performance (10 problems)
- Written explanations to 5 problems
31Meteorology Knowledge
32Weather Prediction
33Use of PGF and Coriolis Principles in Explanations
34Other Heuristics used by Naïve Participants.
- Pressure Temperature association.
- West to East movement of air.
- Temperature area dominance.
- Temperature proximity.
35Conclusions
- Knowledge of principles not sufficient for making
correct inference. - Naïve heuristics still present after learning
correct principles
36Experiment 2c Novices
- Why do Novices (Geography students) fail to make
the correct inferences when they have the
relevant knowledge? - Considered 2 possibilities
- They fail to activate the relevant information
- They are not able to make necessary chain of
inferences, that is, - Infer wind direction from pressure
- Infer temperature change from wind direction
37Method
- 17 students in an undergraduate meteorology class
- 8 participants received a hint
- 9 participants did not receive a hint
- No guidance Does the area get warmer or colder?
- Arrows first 1. Draw arrows
- 2. Does the area get warmer or colder?
- 3. Explain answer
38Meteorology Knowledge
39Weather Prediction
40Arrows
Reminder What the Experts Drew
41Examples of What Novices Drew
42(No Transcript)
43(No Transcript)
44(No Transcript)
45Arrows Consistent with PGF and Coriolis Effect
94 of answers consistent with arrows
46Conclusions Experiment 3c
- Inference guidance lead to somewhat better
performance - Novices performance was based on heuristics,
however - Sometimes only partial knowledge (e.g., PGF but
not Coriolis) - Sometimes misapplication of rules (e.g. reverse
Coriolis)
47Summary
- Ability to read a graph and possession of
knowledge are not sufficient conditions for
making inferences. - Relevant knowledge not always activated
- Naïve theories sometimes activated instead
- Small significant effects of
- Integrated displays
- Inference guidance (arrow task)
48Practical Implications
- Trainees need to be taught explicitly to make
inferences from graphical displays. - Instruction must confront naïve mental models.
- Displays should integrate the relevant
information and show no irrelevant information.