CoSearch: A System for Co-located Collaborative Web Search

34
CoSearch: A System for Co-located Collaborative Web Search Saleema Amershi, Meredith Ringel Morris

Transcript of CoSearch: A System for Co-located Collaborative Web Search

CoSearch: A System for Co-located Collaborative Web Search

Saleema Amershi, Meredith Ringel Morris

• Search often considered to be a solitary activity

Do People Search Collaboratively?

2

Do People Search Collaboratively?

• 3.8 to 1 student-to-computer ratio in U.S. public schools

• 5000 to 3 person-to-computer ratio in U.S. public libraries

• 10 to 1 student-to-computer ratio in developing world schools

3

• YES!

…but current search engines and web browsers do not support collaborative search.

4

Do People Search Collaboratively?

• People• 2 Librarians• 3 Teachers• 2 Developing world researchers

• Questions• Who collaboratively searches the Web?• Why do they collaboratively search?• How do they currently search in co-located settings?

Interview Study

5

Who Collaboratively Searches & Why?

• Youth, Teens & Students

• Seniors & new immigrants

• People in rural regions of the developing world

• Small business employees

• Pedagogical and social value

• Unfamiliarity with technology

• Resource constraints6

• Drivers control input devices

• Observers make suggestions verbally or through gestures

How Do They Collaboratively Search?

7

• Difficulties contributing

• Controlling drivers may ignore observer suggestions

• Demanding observers may make it difficult for drivers to make contributions

8

Limitations

• Pacing problems

• Scrolling too fast or too slow

• Navigating away from a page too quickly

9

Limitations

• Referential difficulties

• Difficulty referring to on-screen content if situated away from the display

10

Limitations

• Single-track strategies

• No division-of-labor

• Inefficient

11

Limitations

Limitations

• Difficulties contributing

• Pacing problems

• Referential difficulties

• Single-track strategies

• Lack of hands-on learning

• Information loss

12

Design Implications

• Facilitate co-located collaborative search

• Enable distributed control and division of labor

• Encourage collaboration, communication and awareness

• Leverage ubiquitous devices (mice and mobile phones)Related work (Inkpen, 1999; Pawar et al., 2007; Paek et al., 2004;

Ballagas et al., 2005; Mahaney and Pierce, 2003; Han et al., 2000)

13

• CoSearch with multiple mice• Refer to paper

• CoSearch with mobile phones• In this talk

14

CoSearch

15

• Individual color-coded cursors

• Also helps to

• Refer to on-screen content

• Enable hands-on-learning

16

Distributing Control

17

Enabling Contributions

• Color-coded Page Queue

18

Page Queue

• Color-coded Page Queue

• Color-coded Query Queue

• Query by text messaging

19

Query Queue

Reading at Your Own Pace

• Viewing Web pages on mobile phones

• Also enables division of labor

20

Status-quo Limitations CoSearch Features

Difficulties contributing Individual color-coded cursors, Query Queue & query by text messaging, Page Queue

Pacing Problems Viewing Web pages in mobile phones

Referential difficulties Individual cursors controlled by mice or mobile phones

Single-track strategies Viewing Web pages in mobile phones

Lack of hands-on learning Individual input devices (mice and mobile phones)

Information loss Notes regions, summaries

Evaluation Goals

• Assess how well CoSearch enables:

• Distributed control

• Division of labor

• Group communication

• Awareness

22

Participants

• 3 person groups, 12 groups• 21 males, 15 females• 12 - 76 years old• Experienced and non-experienced searchers• Experienced and non-experienced mobile phone users• Friends, siblings, children with parents, adults with

grandparents

23

• Within-subject• 3 conditions: CoSearch, Shared, Parallel

• 2 tasks per condition• One fixed (e.g., “Which state is the birthplace of the

most U.S. Vice Presidents?”)• One group-selected (e.g., planning a trip or group activity)

• Questionnaires, log data, observations 24

Study Design

25

Results

• Communication• CoSearch and Shared better than Parallel (p<.01)

• Collaboration• CoSearch and Shared better than Parallel (p<.01)

26

Communication & Collaboration

• Communication• CoSearch and Shared better than Parallel (p<.01)

• Collaboration• CoSearch and Shared better than Parallel (p<.01)

• Frustration• Observers more frustrated in Shared than drivers (p<.03)

• Experienced searchers more frustrated in Shared than less experienced (p<.01)

• No differences in CoSearch

27

Reduced Frustration

• Distribution of Control• “Submit search topics without having to yell at the person

on the computer”

• ”Have more of a say in what’s going on on screen”

• “Go at my own pace”

• Division of Labor• “We could search many offshoots of the same topic at

once”

• “Input more ideas on how to find the answer”

28

Control & Division of Labor

• Overall • #1 Favorite: Parallel (15 participants)

• #2 Favorite: CoSearch (11 participants)

• #3 Favorite: Shared (7 participants)

• CoSearch better than Parallel for communication collaboration

• CoSearch intended for resource-constrained environments where Parallel is not feasible

• CoSearch better than Shared for distribution of control, division of labor, and reduced frustration

29

Overall

• Awareness• Shared better than CoSearch and Parallel (p<.04)

• Experienced SMS users more aware of group in CoSearch than less experienced users (p<.02)

• Feelings of being ignored• More so in CoSearch and Parallel than Shared (p<.01)

• Only 55.3% of observer queries were executed by drivers

• Only 10.88% of observer Web pages viewed by group

30

Problems with CoSearch

• Usability

• People able to quickly learn CoSearch

• More-experienced searchers found it easier than less experienced (p<.03)

• Technological Limitations

• Lag in WiFi and Bluetooth

• Small screens and keypads

31

CoSearch Usability

Recap

• Interview Study to learn about status-quo co-located collaborative search practices (Shared & Parallel)

• Developed CoSearch to address limitations of current practices

• Evaluated CoSearch against current practices• CoSearch better than Parallel for communication and

collaboration

• CoSearch better than Shared for reducing frustrations and increasing control and division of labor

• Still room for improvement in CoSearch

32

Conclusion

• Shared-computing still prevalent in many scenarios.

• CoSearch enhances the shared-computing experience by leveraging additional devices in the environment.

33

Thank you!

34

Status-quo Limitations

CoSearch Features

Difficulties contributing

Individual cursors, Query Queue & query by text messaging, Page Queue

Pacing Problems Viewing Web pages in mobile phones

Referential difficulties

Cursors controlled by mice or mobile phones

Single-track strategies

Viewing Web pages in mobile phones

Lack of hands-on learning

Individual input devices (mice and phones)

Information loss Notes regions, summaries