Using Information Scent to Model Users in Web1.0 and Web2.0

64
e from: http://www.flickr.com/photos/ourcommon/480538715/ Modeling of Web Users from Web1.0 Modeling of Web Users from Web1.0 to Web2.0 to Web2.0 Ed H. Chi, Principal Scientist and Area Manager Augmented Social Cognition Area Palo Alto Research Center 1 2010-03-20 Utrecht CogModeling

description

This talk summarizes the work I have been doing on modeling user behavior on Web1.0 and Web2.0 systems in the last 13 years Talk given at a workshop on Cognitive Modeling in Utrecht, Netherlands on March 20, 2010.

Transcript of Using Information Scent to Model Users in Web1.0 and Web2.0

Page 1: Using Information Scent to Model Users in Web1.0 and Web2.0

Image from: http://www.flickr.com/photos/ourcommon/480538715/

Modeling of Web Users from Web1.0 to Modeling of Web Users from Web1.0 to Web2.0Web2.0

Ed H. Chi, Principal Scientist and Area Manager

Augmented Social Cognition AreaPalo Alto Research Center

12010-03-20 Utrecht CogModeling

Page 2: Using Information Scent to Model Users in Web1.0 and Web2.0

PARC OverviewPARC Overview Interdisciplinary research

center

Founded in 1970

Spun out of Xerox in 2002

Business model:– Contract research

– Licensing

– Joint ventures

– Spinoffs

2010-03-20 2Utrecht CogModeling

Page 3: Using Information Scent to Model Users in Web1.0 and Web2.0

PARC InnovationPARC Innovation

chartered to create the architecture of information & the office of the future- invented distributed personal computing

- established Xerox’s laser printing business

- created the foundation for the digital revolution

Graphical User Interface

Laser Printing

Ethernet

Bit-mapped Displays

Distributed File Systems

Page Description Languages

First Commercial Mouse

Object-oriented Programming

WYSIWYG Editing

Distributed Computing

VLSI Design Methodologies

Optical Storage

Client/Server Architecture

Device Independent Imaging

Cedar Programming Language

2010-03-20 3Utrecht CogModeling

Page 4: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 4

How do people navigate?How do people navigate? Scan Skim Decide Action

2010-03-20

Page 5: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 5

Ecological ApproachEcological Approach

human-information interaction is adaptive to the extent:

Net Knowledge Gained

Costs of InteractionMAXIMIZE [ ]

Page 6: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 6

InformationInformationEnergyEnergy

Analogy to Optimal ForagingAnalogy to Optimal Foraging

Page 7: Using Information Scent to Model Users in Web1.0 and Web2.0

Information Scent: The TheoryInformation Scent: The Theory

Information Scent is the user perception of the cost and value of information.– Similar to hunters following

animal foot prints.

2010-03-20 7Utrecht CogModeling

Page 8: Using Information Scent to Model Users in Web1.0 and Web2.0

Information Scent: The IdeaInformation Scent: The Idea

cell

patient

dose

beam

new

medical

treatments

procedures

InformationNeed Text snippet

SeesWants

• Spreading activation– Bayesian prediction of relevance of individual

elements

2010-03-20 8Utrecht CogModeling

Page 9: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 9

ibread

jbutter

sandwich

flour

Ai = Bi + WjSji

Activation of chunk i

Base-level activation of chunk i

Activation spreadfrom linked chunks j

Activation depends ona base level plus activation

spread from associated chunks

Bi = log( ) Pr(i) Pr(not i)

Sji = log( ) Pr(j|i)Pr(j|not i)

log likelihood of i occurring

log likelihood of i occurring with j

Base level activation reflectslog likelihood of events in the world.Strength of spread reflects log likelihood of event cooccurrance

Page 10: Using Information Scent to Model Users in Web1.0 and Web2.0

Attacking The ProblemAttacking The Problem Users have information goals, their surfing

patterns are guided by information scent

Two questions– Given an information goal and a starting point

Where do users go? (Behavior)– Given some surfing pattern

What is the user’s goal? (Need)

2010-03-20 10Utrecht CogModeling

Page 11: Using Information Scent to Model Users in Web1.0 and Web2.0

WUFIS: Web User Flow by Information Scent

UserInformation

Goal

Web site

WebPage

content links

Web user flow simulation

Predictedpaths

2010-03-20 11Utrecht CogModeling

Page 12: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 12

InfoScent: How does it work?InfoScent: How does it work?

Start users at page with some goal

Flow users through the

network

Examine user patterns

Scent Values: Probabilities of

Transition

2010-03-20

Page 13: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 13

InfoScent SimulationInfoScent Simulation

document

wordWQR

T

1000000

0010000

0000100

0100000

0000011

0000001

0001000

0101110

0

0

0

0

0

0

1

1 from

toTRS

0269.0269.00000

10731.000212.00

0000000

00001576.00

00000212.00

0731.000001

00000002

1

Weight MatrixQuery

Relevant

Docs

R = Relevant documents

T = Topology matrix

Normalize to Probability

Scent Matrix

2010-03-20

Now with the Scent Matrix, we then perform Spreading Activation.

Now with the Scent Matrix, we then perform Spreading Activation.

3

Page 14: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 14

Proximal Cue WordsProximal Cue WordsGoal: Find words that represent Information

Cues for hyperlinks:

Text of the link itself Words around link.– Lists, Paragraphs

1 2

2010-03-20

Page 15: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 15

Information CuesInformation Cues

If the above two fails,– Content words on the Distal Page– Title Words of the Distal Page

3

2010-03-20

Page 16: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20Utrecht CogModeling 16

Bloodhound ProjectBloodhound Project

Starting Point: www.xerox.comTask: look for “high end copiers”

OUTPUTusability metrics

INPUT

Page 17: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 17

Input Input TasksTasks

Page 18: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 18

Stanford CSStanford CS

Page 19: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 19

ONRONR

Page 20: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 20

Instrumentation: WebLoggerInstrumentation: WebLogger

(BEFORE-NAVIGATE (http://altavista.com/ ) 105.331s 0.100s 951763010 10:36:50) (DOC-MOUSEMOVE (881 122 ) 105.431s 0.100s 951763010 10:36:50) (NAVIGATE-COMPLETE (http://www.altavista.com/)105.632s 0.201s 951763011 10:36:51) (EYETRACKER-SYNC (103 ) 106.242s 0.610s 951763011 10:36:51) (DOCUMENT-COMPLETE (http://www.altavista.com/)106.773s 0.531s 951763012 10:36:52) (SCROLL-POSITION (0 0 759 1181 ) 106.853s 0.080s 951763012 10:36:52) (DOC-MOUSEMOVE (874 123 ) 107.024s 0.171s 951763012 10:36:52) (DOC-MOUSEMOVE (874 123 ) 107.044s 0.020s 951763012 10:36:52) (DOC-MOUSEMOVE (874 123 ) 107.214s 0.170s 951763012 10:36:52) (EYETRACKER-SYNC (104 ) 107.244s 0.030s 951763012 10:36:52) (CHAR (a 874 123 ) 108.125s 2.904s 951763013 10:36:53) (EYETRACKER-SYNC (105 ) 108.245s 1.001s 951763013 10:36:53) (DOC-KEYPRESS (a INPUT ) 108.446s 0.201s 951763013 10:36:53)

Page 21: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 21

User TracesUser Traces

Page 22: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 22

Compare Visitation Compare Visitation DistributionsDistributions For each task, produce a user summary vector

that describes the frequency distribution of page visit over the document space.

For each task, ran Bloodhound and create bloodhound predicted frequency distribution.

Page 23: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 23

ResultsResults Corr.Coeff.

Yahoo REI HivInSite Parcweb

task 1a 0.7528 0.4701 0.6811 0.7394

task 1b 0.7218 0.4763 0.7885 0.8756

task 2a 0.7489 0.9892 0.6671 0.8930

task 2b 0.8840 0.7073 0.6880 0.8573

task 3a 0.7768 0.7321 0.8835 0.7197

task 3b 0.6973 0.6979 0.5660 0.7123

task 4a 0.9022 0.9415 0.8407 0.8340

task 4b 0.9052 0.7600 0.4634 0.9344

• Produced click streams that:• Correlated strongly 1/3 of the time• Moderately slightly less than 2/3 of the

time– Problem: we do not know a priori which

third.

Page 24: Using Information Scent to Model Users in Web1.0 and Web2.0

IUNIS: Inferring User Need by Info Scent

UserInformation

Goal

Web site

WebPage

content links

Web user flow simulation

Observedpaths

2010-03-20 24Utrecht CogModeling

Page 25: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 25

Page 26: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 26

Evaluation of IUNISEvaluation of IUNIS

Procedure:– 10 Path booklets– Single rating sheet with the ten 20-word

summaries. A copy of this rating sheet is attached to each of the 10 path booklets.

– Users are asked to read through each booklet and rate each of the path summaries.

Each summary, 5-point Likert Scale. Which of the ten summaries was the best match.

2010-03-20

Page 27: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 27

Evaluation of IUNISEvaluation of IUNIS

Results:– Matching summary mean = 4.58 (median=5)– Non-matching summary mean = 1.97

(median=1)– Difference highly significant (p < .001)– Best match summary: 5.6 out of 10 (Cohen

Kappa=0.51)

Evaluation yield strong evidence that IUNIS generates good summaries of the Web paths.

2010-03-20

Page 28: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 28

ScentTrails: ScentTrails: Pre-highlight Pre-highlight navigation pathnavigation path

A store that knows your goal. Over 50% reduction in task time.

2010-03-20

Page 29: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 29

Web page with highlighted link Web page with highlighted link anchorsanchors

Partial information goal: “remote diagnostic technology”

62 copies/min.

92 copies/min.Remainder of information goal: “speed >= 75”

Utrecht CogModeling

Page 30: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 30

ScentTrails algorithmScentTrails algorithm

Identify tasty pages Waft scent backward along links

– Loses intensity as it travels

remote diagnostics

copiers

fax machines

other maintenance

. . .

XC4411 XC5001

XC4411 copier

featuresFeatures:

remote diagnostics

. . .

digital copiers color copiers

back

Page 31: Using Information Scent to Model Users in Web1.0 and Web2.0

Utrecht CogModeling 31

Results of user studyResults of user study

0

1

2

3

4

5

6

Scent

Trails

ShortS

cent

sear

ch

brow

se

Task

Co

mp

leti

on

Tim

e (m

inu

tes)

0%

10%

20%

30%

40%

50%

Fra

cti

on

Ab

ov

e 5

Min

ute

s(times capped at five

minutes)10/12 subjects preferred ScentTrails to both searching and browsing

2010-03-20

Page 32: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 32Utrecht CogModeling

ScentIndexScentIndex

Associated Entries underlined in red

Page 33: Using Information Scent to Model Users in Web1.0 and Web2.0

33

ScentHighlightScentHighlight

User first type search keywords: “anthrax symptoms”

Conceptually highlight any relevant passages and keywords

Draw user attention

2010-03-20 Utrecht CogModeling

Page 34: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 34

Utrecht

CogModeling

MethodMethod

Page 35: Using Information Scent to Model Users in Web1.0 and Web2.0

User Study SummaryUser Study Summary Overall, the ScentIndex eBook performed

better against the physical Book. Faster Speed:

– Subjects using the ScentIndex were faster in completing their tasks no matter whether they were experts or novices, F(1,12)=12.96, p<.01.

More Accurate:– Answers that they provided while using ScentIndex

interface were more accurate, F(1,12)=3.991, p=.06.

2010-03-20 Utrecht CogModeling 35

Page 36: Using Information Scent to Model Users in Web1.0 and Web2.0

Poor heuristic

Good heuristic

HeuristicsHeuristics

2010-03-20 36Utrecht CogModeling

Page 37: Using Information Scent to Model Users in Web1.0 and Web2.0

““Hints”Hints”

Solo

Cooperative (“good hints”)

2010-03-20 37Utrecht CogModeling

Page 38: Using Information Scent to Model Users in Web1.0 and Web2.0

Finding a Finding a RestaurantRestaurant

Appropriate for the occasion

2010-03-20 Utrecht CogModeling 38

Page 39: Using Information Scent to Model Users in Web1.0 and Web2.0

Research VisionResearch Vision

Augmented Social CognitionAugmented Social Cognition Cognition: the ability to remember, think, and reason; the

faculty of knowing. Social Cognition: the ability of a group to remember, think,

and reason; the construction of knowledge structures by a group.– (not quite the same as in the branch of psychology that studies

the cognitive processes involved in social interaction, though included)

Augmented Social Cognition: Supported by systems, the enhancement of the ability of a group to remember, think, and reason; the system-supported construction of knowledge structures by a group.

Citation: Chi, IEEE Computer, Sept 2008

392010-03-20 Utrecht CogModeling

Page 40: Using Information Scent to Model Users in Web1.0 and Web2.0

Research MethodologyResearch Methodology

Characterize activity on social systems with analytics Model interaction social and community dynamics and

variables Prototype tools to increase benefits or reduce cost Evaluate prototypes via Living Laboratories with real users

40Utrecht CogModeling2010-03-20 40

Characterization Models

PrototypesEvaluations

Page 41: Using Information Scent to Model Users in Web1.0 and Web2.0

412010-03-20 Utrecht CogModeling

Characterization Models

PrototypesEvaluations

Page 42: Using Information Scent to Model Users in Web1.0 and Web2.0

Two Sides of TaggingTwo Sides of Tagging

Encoding Retrieval

42

http://edge.org

“science research cognition”

http://www.ted.com/index.php/speakers

“video people talks technology”

2010-03-20 42Utrecht CogModeling

Page 43: Using Information Scent to Model Users in Web1.0 and Web2.0

Using Information Theory to Model Social Using Information Theory to Model Social TaggingTagging[Ed H. Chi, Todd Mytkowicz, ACM Hypertext 2008][Ed H. Chi, Todd Mytkowicz, ACM Hypertext 2008]

TopicsConcepts

Users Documents

Tags

T1…TnEncodingDecoding

Noise

2010-03-20 43Utrecht CogModeling

Page 44: Using Information Scent to Model Users in Web1.0 and Web2.0

H(Tag) shows saturation in tag usage H(Tag) shows saturation in tag usage

442010-03-20 Utrecht CogModeling

Page 45: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 46

II((DocDoc; ; TagTag) Mutual ) Mutual InformationInformation

Source: Hypertext 2008 study on del.icio.us (Chi & Mytkowicz)

Page 46: Using Information Scent to Model Users in Web1.0 and Web2.0

Raise in avg. tag per bookmarkRaise in avg. tag per bookmark(note parallel the development in increasing # of (note parallel the development in increasing # of query words)query words)

472010-03-20 Utrecht CogModeling

Page 47: Using Information Scent to Model Users in Web1.0 and Web2.0

482010-03-20 Utrecht CogModeling

Characterization Models

PrototypesEvaluations

Page 48: Using Information Scent to Model Users in Web1.0 and Web2.0

• Synonyms• Misspellings• Morphologies

People use different tag words to express similar concepts.

Social Tagging Creates Noise

2010-03-20 49Utrecht CogModeling

Page 49: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 50

Guide

Web

Howto

TipsHelp

Tools

Tip

Tricks

Tutorial

Tutorials

Reference

Semantic Similarity GraphSemantic Similarity Graph

TagSearch: TagSearch: Use Semantic Use Semantic Analysis to Reduce NoiseAnalysis to Reduce Noise http://mrtaggy.com

Utrecht CogModeling

Page 50: Using Information Scent to Model Users in Web1.0 and Web2.0

MapReduce ImplementationMapReduce Implementation

Spreading Activation in a bi-graph Computation over a very large data set

– 150 Million+ bookmarks

Tags URLs

P(URL|Tag)

P(Tag|URL)

2010-03-20 51Utrecht CogModeling

Page 51: Using Information Scent to Model Users in Web1.0 and Web2.0

Understanding a new area…Understanding a new area…

2010-03-20 52

Characterization Models

PrototypesEvaluations

Utrecht CogModeling

Page 52: Using Information Scent to Model Users in Web1.0 and Web2.0

MrTaggy.com: MrTaggy.com: social search browser with social social search browser with social bookmarksbookmarks

Joint work with Rowan Nairn, Lawrence Lee

Kammerer, Y., Nairn, R., Pirolli, P., and Chi, E. H. 2009. Signpost from the masses: learning effects in an exploratory social tag search browser. In Proceedings of the 27th international Conference on Human Factors in Computing Systems (Boston, MA, USA, April 04 - 09, 2009). CHI '09. ACM, New York, NY, 625-634.

2010-03-20 53Utrecht CogModeling

Page 53: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 54Utrecht CogModeling

Page 54: Using Information Scent to Model Users in Web1.0 and Web2.0

Understanding a new area…Understanding a new area…

2010-03-20 56

Characterization Models

PrototypesEvaluations

Utrecht CogModeling

Page 55: Using Information Scent to Model Users in Web1.0 and Web2.0

Baseline Baseline InterfaceInterface

2010-03-20 57Utrecht CogModeling

Page 56: Using Information Scent to Model Users in Web1.0 and Web2.0

Experiment DesignExperiment Design 2 interface x 3 task domain design

– 2 Interface (between-subjects) Exploratory vs. Baseline

– 3 task domains (within-subjects) Future Architecture, Global Warming, Web Mashups

30 Subjects (22 male, 8 female)– Intermediate or advanced computer and web search skills– Half assigned Exploratory, half Baseline.

For each domain, single block with 3 task types:– Easy and Difficult Page Collection Task [6min each]– Summarization Task [12min]– Keyword Generation Task [2min]

2010-03-20 58Utrecht CogModeling

Page 57: Using Information Scent to Model Users in Web1.0 and Web2.0

Procedure [2 hours]Procedure [2 hours] Prior Knowledge Test 1st Task Domain

– With easy and difficult page collection tasks, summarization and keyword generation task.

– NASA cognitive load questionnaire 2nd Task Domain

– Same battery of tasks and cognitive load questionaire

3rd Task Domain Experimental Survey

2010-03-20 59Utrecht CogModeling

Page 58: Using Information Scent to Model Users in Web1.0 and Web2.0

Experimental Evauation Experimental Evauation [Kammerer et al, CHI2009][Kammerer et al, CHI2009]

Exploratory interface users:– performed more queries, – took more time, – wrote better summaries (in 2/3 domains), – generated more relevant keywords (in 2/3 domains),

and– had a higher cognitive load.

Suggestive of deeper engagement and better learning.

Some evidence of scaffolding for novices in the keyword generation and summarization tasks.

2010-03-20 60Utrecht CogModeling

Page 59: Using Information Scent to Model Users in Web1.0 and Web2.0

The TeamThe Team

2010-03-20 Utrecht CogModeling 61

Page 60: Using Information Scent to Model Users in Web1.0 and Web2.0

Image from: http://www.flickr.com/photos/ourcommon/480538715/

Augmented Social Cognition:Augmented Social Cognition:From Social Foraging to Social From Social Foraging to Social SensemakingSensemaking

Research Vision: Understand how social computing systems can enhance the ability of a group of people to remember, think, and reason.

Living Laboratory: Create applications that harness collective intelligence to improve knowledge capture, transfer, and discovery.

http://asc-parc.blogspot.comhttp://[email protected]

622010-03-20 Utrecht CogModeling

Page 61: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 Utrecht CogModeling 63

Page 62: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 64Utrecht CogModeling

Enhanced ThumbnailsEnhanced ThumbnailsAndrew Faulring, Allison Woodruff and Ruth RosenholtzAndrew Faulring, Allison Woodruff and Ruth Rosenholtz

 

enhanced

plain

Page 63: Using Information Scent to Model Users in Web1.0 and Web2.0

2010-03-20 65

Utrecht

CogModeling

Popout PrismPopout Prism [ [Suh &Woodruff]Suh &Woodruff]

Page 64: Using Information Scent to Model Users in Web1.0 and Web2.0

TagSearch Exploratory FocusTagSearch Exploratory Focus

67

3 kinds of search

navigational transactional

28% 13%

You know what you want and where it is You know what you want to do

Existing search engines are OK

informational

59%

You roughly know what you want

but don’t know how to find it

Difficult for existing search engines

Opportunity

2010-03-20 Utrecht CogModeling