6

0

1

2

3

4

5

642

0

14

12108

Mouse Trackball Joystick Touchpad

Troughput Error rate

Trou

ghpu

t (bp

s)

Erro

r rat

e (%

)

Image by MIT OpenCourseWare.

© Mozilla. All rights reserved. © source unknown.

All rights reserved.

© Microsoft. All rights reserved.

© Microsoft. All rights reserved.

© Microsoft. All rights reserved.

© The Eclipse Foundation. All rights reserved.

© Interface Hall of Shame. All rights reserved.

© Microsoft. All rights reserved.

11

5040

30

20

15

108

654

3

2

2 3 4 5 6 8 10 15 20 30 40 50

Predicted execution time (sec)

Obse

rved

exe

cutio

n tim

e (s

ec)

Graphic EditorsMARK UP

DRAW SIL

Executive SubsystemsAll Subsystems

Text EditorsPOET SOS

DISPED

Image by MIT OpenCourseWare.

High-level models of human-computer behavior

• Developing Theories in HCI– must explain and predict human behaviour in the human-computer system

– must work in a wide variety of task situations

– must work within broad spectrum of system designs and implementations

• Some low-level theories can be used to predict human performance– Fitts’ law: time to select an item with a pointing device

– Keystroke level model: sums up times for keystroking, pointing, homing, drawing, thinking and waiting

• General models that explain human behaviour with machines– Syntactic/semantic model (Shneiderman)

– Stages of interaction (Norman)

– all of psychology!

Syntactic/semantic model of user knowledge

•A high level model of interaction, developed by Ben Shneiderman

Action ObjectAction Object

Task Computer

Semantic Syntactic

1. Syntactic knowledge

•The rules or combinations of commands and signals

– seen as device-dependent details of how to use system

– examples:

• backspace key delete previous character

• tab move to next field in a form

• right mouse button context menu

• grep <word> <file> find word in a file

• control-shift-K search messages (thunderbird)

Syntactic

1. Syntactic knowledge (continued)

User problems with syntactic knowledge

syntactic details differ between (and within!) systems

– little consistency –> arbitrary

hard to learn

easily forgotten

Action Object

Computer

2. Semantic knowledge: Computer conceptsThe meaning behind computer concepts

People learn them by– meaningful learning – demonstrations– explanations of features– trial by error– models, analogies

relatively stable in memory– high level concepts– logical structure– cognitive model produced

usually transferable across computer systems– but not always

But: prefer to concentrate on task, not computer knowledge

Action Object

Task

3. Semantic knowledge: task concepts

•The meaning behind the task concepts

– is independent of the computer– Similar mechanism to computer concepts

•Examples– how to write a business plan

• format concerns• stylistic concerns• paragraph structure, etc.

– creating a budget

What Syntactic/Semantic Model reveals

•Mapping between the 3 items is key

– Task semantics to computer semantics to computer syntax• task semantics: write document• computer semantics: open a file, use editor, save it to disk• computer syntax: select menu items, key strokes for

formatting,...

– Bad mapping: using latex to write document• aside from task semantics, must also know semantics/syntax of:

– text editor– latex– Unix compiling and printing sequence (to typeset and print)

– Good mapping: trashcan to throw away files• must know mouse syntax of selecting and dragging• computer semantics almost analogous to task semantics

Guideline from syntactic/semantic model•Reduce burden to task-oriented user of learning separate computer semantics and syntax

•computer semantics– Use metaphors– hide unnecessary information

•computer syntax– A little learning should go a long way...– As simple as possible and uniformly applicable– Generic commands– Syntax consistent between systems

Four Stages of an Interaction

(a simplified version of Norman’s 7 stages)

1. Form intention

2. Select an action

3. Execute the action

4. Evaluate the outcome

The four stages when performing a task

Physical activity

Execution

ActionSpecification

Intention

Goals

Evaluation

Interpretation

Perception

Mental activity

expectation

Example task: improve document’s formattingintention-1

intention-2

intention-3

actionspecification

evaluate-1

evaluate-2

evaluate-3

interpretation

intention-4 evaluate-4

look better

block para

.pp->.sp

execution perception

getformatted

output

actionspecification

execution

interpretation

printer

perception

Image by MIT OpenCourseWare.

GoalsPhysicalSystem

gulf ofexecution

Gulf of Execution

•Do actions provided by system correspond to the intentions of the user?

•Gulf:– amount of effort exerted to transform intentions into selected and

executed actions•Good system:

– direct mappings between Intention and selections

GoalsPhysicalSystem

gulf ofevaluation

Gulf of Evaluation

•Can feedback be interpreted in terms of intentions and expectations?

•Gulf:– amount of effort exerted to interpret feedback

•Good system:– feedback easily interpreted as task expectations

Bridging the Gulfs

GoalsPhysicalSystem

execution bridgeintentions

action specificationsinterface

mechanism

evaluation bridge

interpretationsinterfacedisplay

evaluations