Revisitation Patterns and Disorientation

1
Revisitation Patternsand Disorientation Eelco
Herder Session on Empirical Studies ABIS
Workshop 2003 Karlsruhe
PALS
Research Partners
anywhere
2
Presentation Overview
Introduction Prediction of Disorientation Descript
ion of Pilot Study Experimental Results Discussion
Revisitation Patterns and Disorientation
3
Becoming Disoriented in a Web Site
Symptoms
Where am I How did I come here Where can I go to

• Due to
• non-linear structure of web sites
• lack of context information
• unfamiliarity with site or domain

Introduction
4
Context Information Prevents Disorientation

• Contextual links
• e.g. menus, indexes and site maps
• define users spatial or temporal context
• are needed to understand the site structure
• differ from site to site
• Associative links
• embedded in the content
• interlink semantically related concepts
• do not expose the site structure

Contextual Navigation Support
5
Presentation Overview
Introduction Prediction of Disorientation Descript
ion of Pilot Study Experimental Results Discussion
Revisitation Patterns and Disorientation
6
Task Efficiency and Disorientation
Users may be exploring the site or looking for
specific information. Users that are exploring
a site take some disorientation for granted,
similar to tourists in a city center. We want to
find out when people perceive their sense of
disorientation as a problem.
Goal Directed
Exploring
Navigation Patterns
7
Site Graph and Navigation Graph
Web sites are commonly modeled as directed
graphs User navigation paths can be seen as an
overlay of the site graph
1
2
3
5
4
6
Prediction of Disorientation
8
Characteristics of User Navigation
Navigation graphs describe browsing behavior Path
length and linearity Revisitation (cycles and
clustering) Analysis of navigation patterns
helps in answering questions Is the user
exploring or is he lost? Does the user understand
the site structure?
Prediction of Disorientation
9
Revisitation and Lostness
Contradictory experimental results found in
literature P.A. Smith Users who cannot find
desired information are lost. Lostness measure
based on number of revisits the more revisits,
the more likely the user is lost J.E.
McEneaney Well-performing users return often to
navigational landmarks. The more linear the path,
the worse the performance
Prediction of Disorientation
10
Presentation Overview
Introduction Prediction of Disorientation Descript
ion of Pilot Study Experimental Results on
Revisitation Discussion
Revisitation Patterns and Disorientation
11
Experimental Design (1)
Objectives Find relations betweennavigation
patterns and user satisfaction Subjects 30
randomly selected Dutch students(15 from
Utrecht, 15 from Twente) Session Length Complete
session lasted 2.5 hour Navigation session lasted
only 40 minutes
Description of Pilot Study
12
Experimental Design (2)
Data Gathering Methods Questionnaires (motivation
al factors, expertise, evaluation) Performance
tests (working memory, spatial capacity, episodic
memory) Observation and interview Thinking
aloud protocols (recorded on tape) Navigation
data capturing (Scone)
Description of Pilot Study
13
Web Sites Used and Example Tasks
What are the definitions of personal budget and
financial goal as they are presented within this
website?
Calculate how much the real value of 7.826 will
decrease in 5 years given an annual inflation of
3.
Description of Pilot Study
14
Access Tracking Software
path visualizations site metrics navigation
metrics inference methods
Analysis software
Graph Package
Description of Pilot Study
15
A Navigation Graph
Description of Pilot Study
16
Presentation Overview
Introduction Prediction of Disorientation Descript
ion of Pilot Study Experimental Results Discussion
Revisitation Patterns and Disorientation
17
Revisitation Patterns

• Percentage of Revisits
• About 40 for personal finance tasks,about 25
  for browsing on Amazon
• (Tauscher revisitation rate of 58 on long term)
• Use of Back Button
• On average 9,2 of navigation actions.
• Large differences between sites (3,3 15,8 -
  7,8)
• User tendencies are comparable

Experimental Results
18
Revisitation I will find my way back later
Average Connected Distance (ACD) The expected
length of the path betweenany connected two
pages x and y in the navigation graph. What Does
It Mean? A higher ACD indicates that users will
visit more pages before they return to a certain
pageby a direct link, instead of by multiple
clicks on the back button ACD and working memory
positively correlated
Experimental Results
19
ACD vs Number of Revisits
I will find my way back later means less
revisitation
Experimental Results
20
Return Rate
Return rate The average number of times that a
user will revisit a page. Calculated by
averaging the number of visits on all pages that
have been visited at least twice. What Does It
Mean? A higher return rate means that users
returnto a specific set of pages, most likely
with a specific goal in mind.
Experimental Results
21
ACD, Return Rate and Disorientation
People with high Average Connected Distance have
low revisitation rates and low back button
rates They also make more intensive use of
landmarks (once a page is revisited once, they
come back more often) Intensive use of landmarks
is negatively correlated with disorientation.
Experimental Results
22
Average Time Spent on Pages
Median View Time The average time a user spends
on a page while browsing. In contrast, the
average view time is overly influenced by a small
set of high content pages. Our Findings Users
who feel disoriented, spend more time on pages
while browsing.The correlation is not very
strong, though.
Experimental Results
23
Presentation Overview
Introduction Prediction of Disorientation Descript
ion of Pilot Study Experimental Results Discussion
Revisitation Patterns and Disorientation
24
Revisitation Patterns and Disorientation
Combined measures on user navigation help in
predicting users becoming disoriented. Revisitati
on patterns are more accurate thanthe amount of
revisitation. We are still far away from
perfection.Most likely uncertainty cannot be
eliminated completely.
Discussion
25
Leveraging Lostness

• Even with slightly imperfect prediction of
  disorientation it is still possible to offer
  adaptive navigation support.
• Examples of adaptive contextual navigation aids
• personalized site maps
• visualization of navigation history
• direct guidance
• link annotation
• In order to do that, we must be able to
  distinguish between the various kinds of
  disorientation.

Discussion
26
Limitations of this Study
The site structure has not been taken into
account, due to its complexity.We have evidence
that the structure is an important factor. We
havent yet been able to analyze the relations
between the various measures in detail. This will
involve time-consuming visual analysis, combined
with more specific tasks and more detailed
evaluation.
Discussion
27
Revisitation Patterns and Disorientation Time
for Questions email herder_at_cs.utwente.nl
PALS
Research Partners
anywhere