Adoption and Use of a Presence/Chat Application in Globally-Distributed Software Development
description
Transcript of Adoption and Use of a Presence/Chat Application in Globally-Distributed Software Development
Adoption and Use of a Presence/Chat Application in Globally-Distributed
Software Development
James HerbslebInstitute for Software Research, International
School of Computer Science1321 Wean Hall
Carnegie Mellon UniversityPittsburgh, PA, USA 15213
+1 412 [email protected]
2
Research Team
• University of Michigan– Tom Finholt– Mark Handel
• Carnegie Mellon University– Alberto Espinosa
• Bell Labs Research– David Atkins– David Boyer– James Herbsleb– Stacie Hibino– Audris Mockus– Dewayne Perry– Larry Votta– Graham Wills
3
Nuremberg
NapervilleSwindon
MalmesburyChippenham
Bangalore
Dublin
Paris
HilversumHuizen
Brussels
Columbus
Tools
Rear View Mirror
CalendarBot
Experience Browser
TeamPortal
Models ofDevelopmentHow to distribute work
across global sites.
BestPractices
Design
Code
Test
Planning Travel xxxxxxxxxxxxxxxxx xxxxxx
Establishing Liaisons xxxxxx xxxxx xxxxxxxxxxxxxx xxxxx xxxx xxxxxxxxxxxxxxxx
Building Trust xxxxxxx xxxxx xxxxxxxxxx
Communication Etiquette xxxxxxxxxxxxxxxxxxx xxxxxxxx xxxxxx
Preventing Delay xxxxxxxx xxx xxxxxx xxxxxxxxxxx xxxxxx xxxxxxx
Using Commercial Tools xxxxxxxxx xxxxxxx xxxx xxxxxxx xxxxxx xxxxxxResearch
TeamResearch
Team
Empirical Studies
New ProductsGlobal Development
Solutions
Collaboratory Project
4
Collaboration Over Distance
• Work items split across sites take much longer
• Many fewer communication paths across distance
• Much less frequent communication across distance
• Less comfortable communicating with remote colleagues
• Much more difficult to identify distant colleagues with needed expertise
• Much more difficult to communicate effectively
• Less likely to perceive themselves as part of the same team
• Common view of priorities – no difference
5
RVM Application Background
• Desire to increase informal communication, context information
• Hoped to create more communication “openness”
• Wanted to increase feeling of “teamness”
• First step: low fidelity prototypes– Users unfamiliar with interactive text– Privacy concerns – surveillance tool?
6
Chat, IM, Presence AwarenessRear View Mirror
GroupChat
Presence Viewer
Option to log in at machine startup.
7
Wireless1 Organization
Swindon
Nuremberg
8
Data Collection
• Study covered 17 months– March 2000 – initial introduction– August 2001 – server shut down
• Qualitative data– Twenty interviews– Two focus groups
• Data from usage logs– Who logged in each day– Group Activity– Group chat messages (not IM messages)
9
Initial DeploymentPlanning, Training
• Targeting Key Users– Identifying key cross-site pairs – Want to achieve critical mass as quickly as possible
• Installation and Training– Hour-long sessions with each user– Installation of RVM (and additional tools) with training and
quick reference cards– Initial training done by two-person teams at each site– Intensive one-week push, followed by one week of follow-up– Trained 15 users in England, 15 in Germany– E-mail and phone support afterwards
10
Initial DeploymentAdoption Curve -- Wireless 1
Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug0
10
20
30
40
50
60
70
80
11
Initial DeploymentUse of RVM – Wireless 1
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug
“I Love You” Virus Server Crash Holidays
Reorganizations
12
Adoption
Active users (rolling average)
0.0
1.4
2.8
4.2
5.6
7.0
8.4
9.8
Systems Engineering
Quality
Test
Management 1
Month 1 2 3 4 5 6 7 8 9 10 11 12
13
Initial DeploymentAdoption Issues
• Reliability, usability
• Teams -- key unit for adoption
• Some users -- quick look, duck out style of use
• Individual training sessions – little visible activity
• Privacy versus setup time– Individual permissions very unwieldy
14
Retooling, Rethinking
• Changes in RVM– Bug fixes, testing– Group chat persistence– Group-based security
• Deployment – Team focus
• Targeting users• Training and setup
– Learn how to collaborate, not just how to use the tool
15
Network Team
Columbus
Cary
NapervilleDenver
16
Wireless2 Teams
Dublin
Columbus
17
All GroupsUse of RVM
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug
Wireless 1 (85) Wireless 2 (22) Network (7)
“I Love You” Virus Server Crash Holidays
Reorganizations
18
What Is It Good for?A Pessimistic View . . .
Messages that pop up on screen at an inopportune moment (sometimes from the next cubicle) are destroying workers' concentration. Thoughtless text scrawled and sent in haste can spark online arguments. And in some offices, the question of who is privileged enough to receive certain instant messages is creating the kind of tortured pecking order last seen in high school.
Slatalla, M., The office meeting that never ends. New York Times, Sept. 23, 1999.
19
“Bursty” Chat
250
200
150
100
50
0
Messages and Logins per Day
Time
Chat
Logins
20
Chat Groups
Group Number of Locations
# Chatters (Total Members)
Total messages
Time period (days)
Quality 2 5 (5) 175 235
Research 6 17 (21) 2908 239
Test 1 3 5 (8) 146 224
Kn. Mgt. 3 7 (7) 124 46
Test 2 2 4 (4) 111 116
Arch. 4 4 (8) 104 113
At least 30 days of chatAt least 100 messages
21
Chat CategoriesAvailability Negotiating availability, either now or in the very near future
Non-work topics
Specific non-work content, e.g,. cars, fishing, sports
Work Anything that relates to specific work tasks, processes, or planning; general discussions of business or economics related to work, and discussions about the use of RVM itself.
Greeting Hello, etc., also references to weather or health intended primarily as a greeting. Also, ”closings" such as "Bye!"
Humor Comments clearly intended to be primarily humorous
Other Anything that cannot be categorized elsewhere
Cohen’s kappa = .88Found no flaming.
22
Percentage of Message Type by Group
0%
10%
20%
30%
40%
50%
60%
70%
80%
work availability greeting non-work humor other
Quality
Kn Mgt
Test 1
Architect
Test 2
Research
23
Types of Content by Time of DayPercentage within category
24
Work SubcategoriesCategory % of
MsgsDescription
Technical work
66% Discussion of actual work at hand, e.g., carrying out tests; selecting and defining approaches to the work
Project management
21% Planning the current project, project process issues, project status
Meeting management
8% Planning and running meetings, including the current chat session, as well as future chat and face-to-face meetings, e.g. agenda topics, locations, technology support.
Company 4% Issues affecting the entire company, rather than the current project, e.g.., “You certainly know that discussions [concerning a merger] with [another company] are definitely stopped?”
Other 1% Messages that fit in no other category.
25
Adoption IssuesPerceived Utility
• Interactive text as superfluous
• Is “water cooler” talk real work?
• Users’ perception of causes of multi-site problems– Attribution theory: explain behavior on basis of
personal attributes or situational attributes– Appeared to be tendency to explain “undesirable”
behavior of distant colleagues in terms of personal traits
– If problem is personal traits, is more communication desirable?
26
Groupware’s Critical Mass Dilemma
• Feedback from actual use by groups of real users to get a usable tool– Social and political impacts
• Need a usable tool before you can get critical mass for groups of users– Relatively few innovators, early adopters– Critical mass may not be obtainable
• Possible Solutions:– Extremely tolerant users– Developers use tool– Management pressure– “Progressive” sets of features
27
Research Issues• The role of interactive text in supporting informal
communication in distributed teams– Do remote team members make more person attributions for
“undesirable” behavior?– Does providing context serve to reduce personal
attributions?
• Privacy – How far will group-based model generalize?– What are enduring privacy concerns for different
communities and features?
• Characteristics of RVM Chat that may affect content– Chat participants have known identities (on and off-line)– Chat always “semi-public”– Persists only briefly (unlike b-board)
• Group chat less intrusive than IM?
28
Nuremberg
NapervilleSwindon
MalmesburyChippenham
Bangalore
Dublin
Paris
HilversumHuizen
Brussels
Columbus
Collaboratory Project
Tools
Rear View Mirror
CalendarBot
Experience Browser
TeamPortal
Models ofDevelopmentHow to distribute work
across global sites.
BestPractices
Design
Code
Test
Planning Travel xxxxxxxxxxxxxxxxx xxxxxx
Establishing Liaisons xxxxxx xxxxx xxxxxxxxxxxxxx xxxxx xxxx xxxxxxxxxxxxxxxx
Building Trust xxxxxxx xxxxx xxxxxxxxxx
Communication Etiquette xxxxxxxxxxxxxxxxxxx xxxxxxxx xxxxxx
Preventing Delay xxxxxxxx xxx xxxxxx xxxxxxxxxxx xxxxxx xxxxxxx
Using Commercial Tools xxxxxxxxx xxxxxxx xxxx xxxxxxx xxxxxx xxxxxxResearch
TeamResearch
Team
Empirical Studies
New ProductsGlobal Development
Solutions
We need to finish our preparations for the review!
http://www-spr.research.bell-labs.coCurrent spec.
ConnectIcon
• “Antidote for phone tag”
• Send presence and contact ability to anyone
ConnectIcon
Hi Jim,
We need to talk about the review tomorrow!
Ann Kelly
To check my availability and get my contact information, please click this link:ConnectIcon from Ann Kelly
26 hours ago
23 hours ago
3 days ago
20 hours ago
ConnectIcon
10 minutes ago
Currently in use
Busy