Presented by Kurt Luther (Postdoctoral Fellow, Carnegie Mellon University) at the 2013 ACM Conference on Computer Supported Cooperative Work (CSCW 2013) in San Antonio, TX.
8. Leadership and
Online Creative Collaboration
• Leadership essential to successful online creative
collaboration
• But leadership is hard, and many leaders aren’t
effective
– Newgrounds: leaders overburdened, only 20% of
collaborative animation projects completed [Luther et al. 2010]
• How can we support leaders of online creative
collaboration?
8
9. Preview
• Background: Why online leadership is hard
• Theory: Redistributing leadership
• Design: Pipeline (collaboration tool)
• Case Study: Holiday Flood
• Implications for theory and design
9
10. Challenges for Leaders
• Many responsibilities
– Planning, problem solving, clarifying, etc.
• Challenges of distributed collaboration
– Leadership at a distance
• Challenges of creative collaboration
– Subjectivity, originality, ownership, completion
• Lack of technological support
– Example: Newgrounds “collab threads”
[Becker 1984; Hinds & McGrath 2006; Luther & Bruckman 2008; Yukl 1997]
Background / Theory / Design / Case Study / Implications 10
12. Dealing with the Challenges
• How Newgrounds collab leaders
manage the challenges
– Simplify projects
• Top-down leadership styles
• Minimize interdependence
– Work very hard
• Still, many leaders become
overwhelmed, collabs fail
[Luther & Bruckman 2008; Luther et al. 2010]
Background / Theory / Design / Case Study / Implications 12
13. Research Question
How can we ease the burden on leaders of online
creative collaboration and help them organize
more successful, complex, creative projects?
Design a technology, Pipeline, that helps
leaders redistribute leadership to (1) other
members and (2) the technology itself.
Background / Theory / Design / Case Study / Implications 13
14. Distributed Leadership
• Leadership roles can be separated from
leadership behaviors [Thorpe et al. 2011]
– Only formal leaders hold leadership roles
– Any member might perform leadership behaviors
• Leadership can be distributed in traditional orgs
[Bolden 2011], online collaborations [Zhu et al. 2012]
Design implication: Help non-leaders
perform some leadership behaviors
Background / Theory / Design / Case Study / Implications 14
15. Distributed Cognition
• System-level view of cognition
– Cognitive processes can be
distributed across social groups,
artifacts, time
• Systems of humans and artifacts
fly airplanes [Hutchins 1995], fight
vandalism in Wikipedia
[Geiger & Ribes 2010]
Design implication: Technology itself could
perform some leadership behaviors
Background / Theory / Design / Case Study / Implications 15
17. Pipeline
• Web-based collaboration tool
for creative projects
• Augments, rather than
replaces, existing communities
– Newgrounds, Georgia Tech, etc.
• Released as free, open source
software
Background / Theory / Design / Case Study / Implications 17
24. Studying Pipeline Usage
• Case study: Holiday Flood (HF)
– 6-week Newgrounds art collab
• Data sources
– Pipeline server logs (1100+ events)
– Discussion on Newgrounds forums (140+ posts)
– Interviews with 5 most active members (10.5 hours)
Background / Theory / Design / Case Study / Implications 24
25. Leadership in Holiday Flood
• Five behaviors for managing the work [Yukl 1997]
– Planning
– Problem Solving
– Clarifying Roles & Objectives
– Informing
– Monitoring
Background / Theory / Design / Case Study / Implications 25
26. 12 Drummers 6 Geese Laying
Drumming
11 Pipers Piping 5 Golden Rings
10 Lords Leaping 4 Calling Birds
9 Ladies Dancing 3 French Hens
2 Turtle Doves
8 Maids Milking
A Partridge in a
Pear Tree
7 Swans Swimming
26
28. Planning
• Task system helped HF’s “Holiday Flood was all
planned and plotted.
leaders develop and share It’s the reason we
“action plans” with members needed Pipeline. I
doubt it would have
• Pipeline’s tools encouraged worked out any other
way.” –Renae
leaders to plan a more
complex, ambitious collab
– Multiple phases, deadlines
– High interdependence
Background / Theory / Design / Case Study / Implications 28
30. Problem Solving
• Dealing with emergencies, ZaneZansorrow’s last-
minute contribution
unexpected situations requiring “saved the whole
immediate attention [Yukl 1997] project, basically.”
–Gabriel
• In HF, Pipeline provided
awareness, access to
resources to help leaders and
non-leaders solve problems
– Poster: Multiple members iterate
on big task
– Emergency drummers: Replacing
a drop-out days before deadline
30
32. Informing
• Gathering, sharing relevant Informing was “a clear
sign that the leader is
info with members [Yukl 1997] very much motivated
• Pipeline’s activity feeds with the project.”
–Robert
offloaded some informing tasks
(e.g. group awareness) from
HF’s leaders
• Feeds redistributed informing
to members, not just software
• Human informing provides
motivation, not just information
Background / Theory / Design / Case Study / Implications 32
34. Implications
• Implications for theory
– Distributed leadership and distributed cognition
provide complementary perspectives on designing for
redistributed leadership
– Leadership in online creative collaboration can be
redistributed to encourage more
successful, complex, creative projects
Background / Theory / Design / Case Study / Implications 34
35. Implications
• Implications for design
– Assume influential formal leaders, but help them
redistribute leadership beyond themselves
– Beware of unintended human costs of redistributing
leadership to software
– Support a variety of leadership styles
Background / Theory / Design / Case Study / Implications 35
36. Thank You
• Co-authors: Amy Bruckman, Casey Fiesler
• Pipeline developers: Boris de Souza, Chris
Howse, Kevin Ziegler, Joe Gonzales
• Georgia Tech ELC Lab
Try Pipeline!
• CMU Social Computing Lab http://pipeline.cc.gatech
.edu/
• Pipeline users, beta testers
Fix Pipeline!
• Newgrounds community https://github.com/kluth
• NSF CreativeIT-0855952 er/Pipeline/
36
Editor's Notes
Thank you all for coming. Today I’m going to be talking about redistributing leadership in online creative collaboration. This is joint work with my colleagues Casey Fiesler and Amy Bruckman at Georgia Tech.
I want to start by showing you a collaborative art project called Holiday Flood. Each of these squares represents an original artwork by a different artist or pair of artists. There are 24 in total, two for each of the 12 days of Christmas. Here are a few examples…
Six Geese a-laying, by Re2deemer
Three French Hens by Kash & ComicDeath
And Two Turtle Doves, by TurkOnAStick
The whole project was organized with Pipeline, a free software tool we built to support leaders in online creative collaboration.
But the real credit goes to the nearly 30 artists from more than 12 countries who worked on Holiday Flood. They were led by two effective and influential leaders, (CLK) Renae and Robert, who came up with the idea, recruited artists, divided up the work, and helped keep everyone on schedule.
Holiday Flood provides just one example of the importance of leadership in online creative collaboration. When leaders are effective, these projects can be remarkably successfulSmall example like HF, or world changingWikipedia, OSS, Star Wars Uncut, Polymath projectBut not all leaders effective because leadership is harde.g. NewgroundsWhat I want to talk about today: how to support leaders?
To address this, I’ll first discuss why we think online leadership so hardNext, theory of redistributing leadership to make it easierPipeline, a tool we designed based on this theoryA case study of HF, the proj you saw earlierAnd finally discuss imps for design
Many challengesFirst, they simply have a lot to doDistribcollab – group awareness, trust, coordination often more complicatedCreative collab –artistic/expressive domains – special challengesLack of support – exacerbates the other challenges.
Newgrounds – collabs organized on forumsDesigned for conversation not complex collaboration
Of course, some collabs succeed despite these challengesIn the Newgrounds animation community, two main approaches Simplify Work really hard
We know that leaders face many challenges and frequently become overburdened, causing projects to fail. In this research we ask, how can we ease that burden and foster more successful, complex, and creative collabs? (CLK)Our approach is to design technology that redistributes leadership to other members and technology.I’ll describe the theories behind these ideas.
The idea of non-leaders performing leadership may seem strange, but it’s not new idea – it goes back to interesting work in 1990s. Theory is called DLBasic idea: roles separate and behaviorsSeen examplesDL usually descriptive, but we explore if we can design for variety of distributions
The other part of our approach is to redistribute leadership to the technology itself. Once again, we can draw upon theoryAs many of you know, Dcog says cognition can be distrib across groups, artifacts, timeLet environment do some of the workWe’ve seen systems of humans and technologies fly airplanes, even fight vandalism on WPMaybe leadership could be distrib in similar way
From DL and DCog, we derived design inspiration for how we might ease the burden on leader.These insights Informed the design of Pipeline
Web-based collab toolCross between a project management tool and a peer production platform
Standard features:User profiles
Public and private projects
Task w/ contributions
AND Threaded discussions for higher-level issues.
One important way that Pipeline helps redistribute leadership across the group w/ its trusted member system. (CLK) In Pipeline, Trusted members can do certain kinds of articulation work, like creating and leading tasks, or editing high level project settings.Creators can trust just themselves, the benevolent dictator model, or they can trust many members, project works much like a wiki.
Pipeline also helps redistribute (tedious) leadership across technology.(CLK) For example, leaders spend a great deal of time managing files, so Pipeline supports file uploads of most kinds of media, automatically generates thumbnails and streaming versions, and links together multiple iterations.(CLK) Leaders also expend a lot of effort maintaining group awareness, so Pipeline automatically captures member activity and displays it in context-sensitive feeds.
We launched Pipeline in late 2011. Since then, our Pipelines have drawn more than 450 registered members and 100 projects, including animations, games, videos, artworks, and writing projects.We focused on a case study of one project, an artistic collaboration called Holiday Flood, to paint a detailed portrait of how Pipeline affected leadership “in the wild.” For this case study, we collected and analyzed three sources of data: (1) server logs from the Holiday Flood Pipeline project; (2) discussion of Holiday Flood in the Newgrounds Art Forum; and (3) in-depth interviews with the five most active members of Holiday Flood. Going forward, we use the real names of participants who requested it.
I’ll discuss leadership in terms of Yukl’s five behaviors for managing work.We saw examples of Pipeline redistributing all five, but for time’s sake I’ll only touch on 3 boldfaced. Many more details on the others in our paper
If you’ll recall, HF was a collaborative artwork organized on Pipeline with artists recruited from the Newgrounds community. The goal was to create 24 artworks for the 12 days of Christmas.
The two formal leaders, Renae and Robert, started by creating these tasks where artists could claim their artwork and post works in progress.
We found that Pipeline facilitates planning in at least two ways. First, the task system helped leaders develop concrete “action plans” and share them with artists. This would have been much trickier on the Newgrounds forums. More generally, Pipeline’s features emboldened Holiday Flood’s leaders to plan a much more complex and ambitious project than they had tried before. Unlike previous floods, Holiday Flood had multiple phases and deadlines, and many tasks were interdependent. If even one artist dropped the ball, “the whole thing would’ve been ruined.”(read Renae quote)
In addition to the 24 artworks, HF also needed a large, elaborate Season’s Greetings poster to tie everything together.Claimed by Gabriel, but too much.Robert iterated, but then got busy.Renae finished it up
Poster – example of how Pipeline distributed leadership behavior of problem solvingHF provided awareness, accesse.g. awareness and access to poster files
Another example - When one member suddenly dropped out just four days before first artwork (12 Drummers Drumming) was due, the leader posted a request for help. Within a day, another member, Zanezanesorrow, responded and drew up a replacement artwork. Gabriel recalled, “it saved the whole project, basically”.
Finally, we found evidence that Pipeline redistributed informing in HF, but not exactly how we expectedIdea: inspired from DcogLeaders offload some of the work of group awareness to the software, via activity feedsGenerally worked well, but 2 surprisesFeeds didn’t just push informing to software, but also little bits to members instead of leader informing all of new member, new member must click “Join”
Unexpected benefits of human informing leader “checks off” task despite auto-check to show they paid attention motivates members NOT redundant
I’d like to finish by discussing some theoretical and design implicationsDL and DCog complementary, but had to push bothDL tells us how to distribute leadership across people, but little to say about techDCog suggests ways to include people and artifacts in a system, but doesn’t speak to leadershipIntegrating both gave us what we needed to design PipelineFrame leadership dist as something to fix, and it can be fixedWorked well for HF, but need to study more projects, domains to understand strengths/drawbacks of this approach
Assume formal leaders Robert, Renae influential, but got help Pipeline uses trust system, other ways possibleUnintended costs think we’re redistrib to software, but actually more work for non-leaders unintentionally stripping out important benefits of human participation (e.g. motivation)Variety of leadership (not one size fits all) encourage experimentation enable more successful, complex, and creative projects.
With that, I’d like to thank…CollaboratorsPipeline teamColleagues at GT, CMUPipeline and Newgrounds communitiesNSFAnd I’ll take your questions