Choosing the Best Path: Techniques for Assessing and Improving Information Scent

1
Choosing the Best Path Techniques for Assessing
and Improving Information Scent

• Jason Withrow
• Internet Professional Instructor
• Washtenaw Community College

2
Overview

• Introducing Information Scent
• Developing Good Scent
• Assessing Scent Quality
• Supporting Scent

3
Introducing Information Scent

• What is Information Scent?
• How do you get from here to there in an
  information space?
• Visual and verbal cues guide the way
• Looking for a movie DVD on a website?
• Choose the global navigation link labeled Films
  before the links labeled Music and Software
• But what if you want a movie soundtrack?

4
Theoretical Foundations

• Information Foraging Theory
• Humans are informavores
• We forage through information spaces, looking
  for the best information, with parallels to how
  animals hunt
• We move on when a certain threshold is reached,
  either high or low
• Poor information scent could provide the
  justification for moving on

5
Research Foundations

• Work on information foraging and information
  scent started in the 90s and has primarily been
  by researchers, many in the area of information
  visualization
• Xerox PARC (Peter Pirolli, Stuart Card, Ed Chi,
  among others)
• Additional researchers (e.g., George Furnas, UM
  School of Information)

6
Web Information Scent

• Scent applies to any information space
• Our focus is on the web
• Both browsing and search
• With browsing, good labeling is vitally important
  in getting users to the desired destinations
• With search, the goal is to get the user to the
  right neighborhood on the website

7
Understanding Information Scent

• Each label on a website has a semantic
  relationship with the links to which it leads
• Think of the top-level label as carrying a
  residue of the lower-level labels. This residue
  is the scent we follow.
• Careers carries a strong, distinct residue for
  the Open Positions and Employee Benefits
  subnavigation links.

8
The Perils of Unclear Scent

• Where do I go to buy their software? Products
  or Store?
• Scent should be strong and distinct, leading
  users both towards and away from certain sections
  of a website.

Source http//www.barebones.com
9
Scent and Information Processing

• Choosing among information scents seems to
  involve preconscious processing
• Scent draws on our existing semantic networks,
  vast numbers of nodes (with one node per concept)
  interconnected in various relationships

10
Semantic Networks
Tie
Fire
Pants
Shirt
Red
Blue
Clouds
Plants
Sky
Green
Airplane
Grass
11
Spreading Activation

• Activation of one node spreads down the paths to
  related nodes, in a ripple effect
• As the activation spreads further from the
  source, it decreases in strength
• Distance of nodes from one another, as well as
  the weight (strength) of the connection, is based
  on how closely related they are in your experience

12
Spreading Activation and Scent

• The labels chosen for links activate these nodes
  and cause the spreading activation
• We choose the link label with the strongest
  relationship to what we are seeking, based on
  what we have encoded in our semantic networks

13
Determining Good Scent

• Browsing The user starts at the home page and
  arrives at the desired information simply by
  choosing the best link at each level of the
  site.
• Search The user searches and either finds the
  desired information or arrives at a page where
  local navigation conveys sufficient scent to
  reach the goal.

14
Three Indicators of Poor Scent

• Indecision (Which path to take? More than one
  looks like a possibility.)
• Frustration (None of these look good!)
• Confusion (What does this word even mean?)

15
Why It Matters to Get Scent Right

• Saved time
• Saved patience
• Increased productivity
• Increased satisfaction
• Increased usability

16
Developing Good Scent

• When designing/redesigning a website, aim for a
  broad, shallow structure
• Why breadth over depth?
• Top level labels must provide scent for all
  levels further down
• Degree of labeling ambiguity corresponds to
  degree of scent ambiguity

17
The Value of Breadth
CHI 2001
CHI 2002
- Expanded breadth (11 to 14 links) - Better
scent Call for Participation --gt Submissions,
Volunteering Introduction Overview --gt
FAQ, Conference Overview Presenters --gt
For Presenters - User testing supported changes
Sources http//www.acm.org/chi2001,
http//www.acm.org/chi2002
18
Establishing Breadth Card Sorting

• Exploratory card sorting can be helpful
• Provide users with the content pieces and have
  them sort the content into related groupings,
  then label the groupings
• This is useful primarily for establishing breadth
  and site structure (hierarchy)
• User-supplied labels can sometimes be good at
  conveying scent

19
Exploratory Card Sort Process

• 1. Orient the user (What is the site? Task?)
• 2. The user groups related cards into piles
• 3. The user assigns one label to each pile
• 4. Can the piles be subdivided further?
• 5. Label each of the smaller sub-piles
• 6. Sometimes further subdivision is needed
• 7. Record the groupings and labels
• 8. Repeat with another user

20
Card Sorting Example Election Website

• Candidates bio Election issues Press releases
• Campaign events Speeches Campaign donations
• Media coverage Campaign timeline Voter
  registration
• Website feedback Newsletter
  Endorsements
• Ask the candidate a question

On-the-road journal
Candidates record and accomplishments
Let a friend know about this website
Candidate comparison
Privacy policy Related links Sitemap
Frequently asked questions
Volunteering Campaign staff and openings
21
Create Primary Groups
Candidates bio
Privacy policy
Candidates record and accomplishments
Website feedback
Campaign staff and openings
Ask the candidate a question
Sitemap
Related links
Press releases
Election issues
On-the-road journal
Candidate comparison
Campaign timeline
Frequently asked questions
Volunteering
Media coverage
Endorsements
Speeches
Voter registration
Newsletter
Campaign donations
Campaign events
Let a friend know about this website
22
Label Primary Groups
Privacy policy
About the Candidate
Website feedback
Candidates bio
Sitemap
Candidates record and accomplishments
Related links
Campaign staff and openings
Ask the candidate a question
News Events
Press releases
On The Issues
On-the-road journal
Getting Involved
Election issues
Campaign timeline
Candidate comparison
Volunteering
Media coverage
Endorsements
Frequently asked questions
Speeches
Voter registration
Newsletter
Campaign donations
Campaign events
Let a friend know about this website
23
Create Secondary Groups
News Events
Campaign timeline
Media coverage
Campaign events
Press releases
Newsletter
On-the-road journal
Speeches
24
Label Secondary Groups
News Events
Events
In the Media
Campaign timeline
Media coverage
Campaign events
Press releases
Newsletter
News from the Candidate
On-the-road journal
Speeches
25
Analyzing the Data

• Eyeball the data for common groupings and
  number of top level categories
• Use a program for analysis (as well as
  administration of the card sort)
• EZSort/USort
• WebCAT
• Cluster analysis, a statistical technique, is
  useful for identifying groupings

26
Breadth and Similarity Matching

• User rates on a scale of 1-10 the similarity of
  every possible pairing of content cards
• Cluster analysis creates the groups by crunching
  the numbers and seeing which items are rated as
  being most similar
• No labels are suggested for each cluster of
  content items, but hopefully a clear label
  emerges from examining the groupings

27
Assessing Scent Quality

• Two techniques help in assessing the quality of
  information scent
• Confirmatory card sorting
• User testing

28
Confirmatory Card Sorting

• Conducted after the site architecture has been
  developed
• Asks the question Do users expect to find
  content under the right label?
• If users sort content under the wrong label (or
  cannot place the content at all), that strongly
  suggests scent issues with the current labeling

29
Confirmatory Card Sort Process

• 1. Orient the user (What is the site? Task?)
• 2. Lay out cards with global navigation labels
• 3. User puts content cards under the appropriate
  global navigation label
• 4. Lay out cards with second-level labels
• 5. User subdivides content cards under new
  second-level labels
• 6. Lay out third-level cards and sort further

30
Provide Global Navigation Cards
Getting Involved
About the Candidate
Related Links
Sitemap
On The Issues
News Events
Privacy Policy
Website Feedback
31
First Pass at Dividing Cards
Getting Involved
About the Candidate
Related Links
Sitemap
Candidates bio
Endorsements
Candidates record and accomplishments
Voter registration
Campaign staff and openings
Campaign donations
Ask the candidate a question
Let a friend know about this website
On The Issues
News Events
Privacy Policy
Website Feedback
Election issues
Press releases
On-the-road journal
Candidate comparison
Campaign timeline
Frequently asked questions
Media coverage
Speeches
Newsletter
Campaign events
32
Provide Second-Level Labels
News Events
Press releases
On-the-road journal
Campaign timeline
Media coverage
Speeches
Newsletter
Campaign events
News from the Candidate
Events
In the Media
33
Further Subdivision Occurs
News Events
Events
In the Media
Campaign timeline
Media coverage
Press releases
Campaign events
Newsletter
News from the Candidate
On-the-road journal
Speeches
34
User Testing and Scent

• User testing of information scent tends to work
  best with focused, information-seeking tasks

35
Quantitative User Test Metrics

• Path directness
• determine optimal path and number of clicks
• calculate number of clicks it takes user to reach
  destination and compare
• Path frequency
• which paths are chosen most frequently?
• Time Completion Rate
• Satisfaction

36
Qualitative User Test Metrics

• User comments
• both written and verbal
• Signs of indecision
• hovering back and forth between two global
  navigation links
• Indications of frustration and confusion

37
Supporting Information Scent

• From user testing in particular lots of
  suggestions arise for supporting information
  scent
• These often relate more to interface design
  decisions than to conceptual design

38
Support Options

• Scope indications
• See Also links
• Facet-based browsing
• Scent stress test for search

39
Indicating Scope

• Scope refers to the nature and extent of content
  in a specific part of a website
• Scope helps establish context for top-level
  labels, clarifying their scent
• Scope can be represented through
• Textual description
• List of subnavigation links

40
Text Description of Scope
Link descriptions can also be provided using the
title attribute.
Source http//www.ustransplant.org
41
Scope Links
Source http//dmoz.org
42
Scope and Graphical Rollovers
Source http//www.interlinknetworks.com
43
See Also Links

• Often during card sorting a piece of content can
  potentially go into two piles
• Once a final location for that content is
  determined, provide a See Also link to the
  location from the other part(s) of the site where
  it could have been located
• This will help the users that follow the wrong
  scent initially

44
Standard See Also Links
Source http//www.ehawaiigov.org
45
Contextual See Also Links
Source http//www.sims.berkeley.edu
46
Facet-Based Browsing

• Rather than trying to capture the whole scent in
  one label, another option is to allow browsing by
  facet
• A facet is an aspect or dimension of an object or
  piece of information
• Each facet is a scent trail that can lead to the
  object or information

47
Facet-Based Browsing
Source http//www.wine.com
48
Scent Stress Test for Search

• Users can enter your site at any page (assuming
  the website is not using frames)
• If they enter at a subpage, what scent will there
  be to assist in navigation?
• To support scent at those lower levels, a stress
  test can be performed

49
Stress Test Criteria

• Looking at the subpage, identify the following
  items
• The name of the website
• The title of the page
• The section of the website you are in
• The path from the home page to your location
• Other pages at the same level
• Pages further down from your location

50
Summary

• The value of information scent
• Incorporating scent into conceptual design and
  interface design