Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Toward an Understanding of Low-Overhead, Action-Oriented
Team ScienceMarshall Scott Poole
Research Team
• Iftekhar Ahmed• Beth Simpson• Natalie Lambert• Ashley Trudeau
Collaboration
• Group decision-making• Conflict management• Team performance• Collaboration technologies• Organizational change• Team Science
Focus• International Virtual Research Organizations– Large labs or centers– Multinational– Multidisciplinary– Dispersed facilities and participants– Diverse participants: Senior scientists, technologists,
postdocs, graduate students, management• Increasingly important vehicles for science– Virtual Astronomical Observatory– AVATOL– HathiTrust Research Center
Challenges for IVROs
– Diversity– Large size– Goal misalignment– Competition– Knowledge integration– Staying organized – Geographical dispersion– High task interdependence in basic work
Question
• How can IVROs promote effective collaborations?
– How can they manage interorganizational alliances?
– How do they promote effective projects?
JLESC
• National Center for Supercomputing Applications (NCSA) • Institut National de Recherche en Informatique et en Automatique
(INRIA)• Argonne National Laboratory (ANL)• Centro Nacional de Supercomputación (BSC)• Jülich Supercomputing Centre (JSC)• RIKEN Advanced Institute for Computational Science (RIKEN)
JLESC• Founded in 2009 by French academic computing
organization INRIA and the National Center for Supercomputing Applications (NCSA)
• Leverage Blue Waters SC at NCSA, let French researchers participate in its development
• Purpose was to bring SCs into contact to engage ideas, advance state of the art
• Criterion for membership: Substantive contribution and willingness to engage in meangingful exchanges with other SCs
Joint Laboratory for Extreme Scale Computing
• Objectives:• Initiate and facilitate international collaborations on
research and state of the art practice topics related to computational and data focused simulation and analytics at scale.
• Address most critical issues in advancing from petascale to extreme scale computing
• Products:• Original ideas• Publications and research reports• Discussion forums• Products (simulations, applications)• Open source software https://jlesc.github.io/
JLESC: Effectiveness
• 75+ interdisciplinary projects• 70+ interdisciplinary publications• 6 external grants• > 100 personnel exchanges• Collaborative learning in cutting edge area• Growth from 2 to 6 institutional members in six
years
Co-authorshipNetworkwithaffiliationdata2009- 2014
Case Study
• Good example of large scale initiative characteristic of “Big”/team science & engineering
• Open and cooperative—wants to understand and improve itself
• Experimental attitude
Case Study (Ongoing)
• Participant observation– Workshops– Governance meetings
• Interviews– Management– Collaboration
• Documents
Two Levels of Analysis
• “Executive Suite”: Interorganizational Alliance
• “Shop Floor”: Collaborative projects
Interorganizational Alliance Level
Challenges Facing JLESC: Interorganizational Level
• Different legal regimes• Competition– Machines– Intellectual property– Personnel
• Different management practices• Funding barriers resulting in resource shortages• Cultural Issues
Types of Interorganizational AlliancesHIERARCHICAL
RELATIONSThrough acquisition or merger, one firm takes full control of another’s assets and coordinates actions by the ownership rights mechanism
JOINT VENTURES Two or more firms create a jointly owned legal organization that serves a limited purpose for its parents, such as R&D or marketing
EQUITYINVESTMENTS
A majority or minority equity holding by one firm through a direct stock purchase of shares in another firm
COOPERATIVES A coalition of small enterprises that combine, coordinate, and manage their collective resources
R&D CONSORTIA Inter-firm agreements for research and development collaboration, typically formed in fast-changing technological fields
STRATEGICCOOPERATIVEAGREEMENTS
Contractual business networks based on joint multi-party strategic control, with the partners collaborating over key strategic decisions and sharing responsibilities for performance outcomes
CARTELS Large corporations collude to constrain competition by cooperativelycontrolling production and/or prices within a industry
FRANCHISING A franchiser grants a franchisee the use of a brand-name identity within a geographic area, but retains control over pricing, marketing, andstandardized service norms
LICENSING One company grants another the right to use patented technologies or production processes in return for royalties and fees
SUBCONTRACTORNETWORKS
Inter-linked firms where a subcontractor negotiates its suppliers’ long-termprices, production runs, and delivery schedules
INDUSTRYSTANDARDS
GROUPS
Committees that seek the member organizations’ agreements on theadoption of technical standards for manufacturing and trade
ACTION SETS Short-lived organizational coalitions whose members coordinate their lobbying efforts to influence public policy making
MARKETRELATIONS
Arm’s-length transactions between organizations coordinated only through the price mechanism
Source:Todova&Knoke,2005
Riskiness of Interorganizational Alliances
• In private sector 50% fail rate• Many non-failures are absorbed back into one
parent (is this really a failure?)• Main benefit consistently reported is for
component organizations…not for IVRO itself…so there are incentives to exploit & free ride
How can vibrant, effective IVROs be maintained?
Alternatives for IVROs
• Association/Conference-Based– e.g., eSocialScience– Convene and develop a community– Collaboration depends on individual initative
• Platform– e.g., International Virtual Astronomical Observatory – Distributed Tools Developed to Meet Standards– Interoperability across international organizations
Alternatives for IVROs
• Centralized Organization– e.g. CERN– Funding in common pool– Legal compact– Centralized management– Project portfolio
• R&D Consortium• JLESC
JLESC: Structure• Consortium of Supercomputing Centers (SCs)
• Each with its own budget, resources, personnel• Each center is run differently• Free to work with whomever they want• JLESC provides a structured “space” for R&D and Learning
• Collaboration Around Set of Topics• Programming models and run time• Numerical algorithms and libraries• Parallel I/O systems and libraries• Resilience and fault-tolerant computing• Performance tools• Applications• Clouds and new architectures
JLESC Structure
• Organizational Structure– Executive Committee-Operating decisions– Steering Committee-High level strategy– Science Committee-Content advice
JLESC: Governance
• Memorandum of Agreement
• Very brief“The lawyers couldn’t agree on anything…We started with 25 pages and by the time they figured out what they could agree on we were down to 3 pages”
• Informally-enforced“An MOU is as good as the people’s intention behind the paper. Nine out of ten MOU’s/interorganizational collaborations like JLESC don’t work.” It depends on developing trust and having someone who makes the collaboration happen
JLESC: Governance
• Relational Contract– Based on trust, reciprocity, norms– Social capital embedded in multiple exchanges that
develop mutual understanding– Partners “continually elaborate on mutual objectives,
capabilities, resources, and tasks” – Presumes concord, not opportunism is basis for
relationships“You have to trust the other will not take advantage”“Of course we are in competition, but the fact that you collaborate even with a competitor improves what you do. “
JLESC
• Resources– Each SC contributes “dues”—a small amount– Each SC makes needed computing resources available– Each SC pays own personnel
• Coordination– Two workshops each year, brings everyone together– Webpage not used for communication but for
information– Linkages through exchange of personnel
Development
• Gradual growth:– INRIA+NCSA– INRIA+NCSA+Argonne NL– INRIA+NCSA+Argonne+Barcelona SC– INRIA+NCSA+Argonne+BSC+Julich SC + RIKEN SC
• Gradual elaboration of structure/governance:– Very small changes from original
• E.g. Steering Committee requested greater formalization and Executive Committee specified one additional role, topic leader
Why Does it Work?
• Modularization reduces scope of challenges to manageable levels
Interorganizational coordination task high uncertainty since SCs do not want to cede autonomy formallySo informal coordination by mutual adjustmentLocal problem solving at SC level or between 2 SCs
• Relatively low formalization allows easy adaptation
e.g. adding new member organizations
“Management by a vision rather than management by a specific plan”
Why Does it Work?• Comfort zone: Capitalizes on general approach in
engineering culture: Let’s focus on immediate problem, define it clear, let’s do something…
• That creates understanding of their work as low in structurability
• Mutual adjustment: Best way to coordinate in a relationship that is moderate to low structurability
• Not resource intensive
• Based on good will and positive contributions
Project Level
Challenges Facing IVROs: Project Level
• Coordination costs of working across organizations/centers
• Different incentive systems across organizations/centers
• Cultural differences in work habits and collaboration
• Occupational divergences
Approaches to Cultivating Collaborations
• Community Based Approaches– e.g., David Straus
Approaches to Cultivating Collaborations
• Team-Based Approaches(team, that is, 2 or more people working together)
– Team creativity (e.g. Hoever et al., 2012; Sawyer, 2006)
– Teamwork (e.g. Salas et al., 2005)
Approaches to Cultivating Collaborations
• Tools-Based Approaches
– Creativity Tools– Thinklets (de Vreede, Briggs, Neiderman, 2008)– Group decision support/groupware/meeting spaces– Fab Labs
Approaches to Cultivating Collaborations
• Unstructured Approaches
– Pick the right people– Put ‘em together– Leave ‘em alone– Hope something comes out of it
– AKA: “Hope and Pray” Based Approach
JLESC Approach
Unstructured…
• Workshops• Projects• Visits
– Organized around areas/themes
– Reliance on old-fashioned human contact, no virtual teams, not much virtual work: Paradoxical that SC people would avoid technologies to support interaction
JLESC: Workshops
• Limited, selective attendance (~200 attendees)• Managers at each SC select people carefully within
thematic areas• Presentations on projects; “stir up interest” in topics• Expectation of collaboration• Occasions for collaboration
– Breaks– Formal sessions– Grad student/postdoc dinner
• Register collaborative projects at conference (form for proposals)
JLESC Workshops
• Probably succeed more due to sheer number of possible collaborations 1302 than due to quality of process of facilitating collaboration
“The workshops have been partly successful. They could possibly be redesigned to make the intent clearer. They have been great in many respects, but could use a clearer focus, a roadmap. Is it a mini-conference? Not really, but if it’s not that.. There’s a lot of talks that there’s a lot of stuff we could do together, but not many collaborations come out of that.”
JLESC Workshops
• Evolution of Collaboration Forums2nd workshop: Just get together and talk3rd workshop: Open mike sessions (ugh)4th workshop: Refined open mike sessions (but no consistent format); Started short presentations5th workshop: Dropped open mike sessions; Refined short presentations (but no consistent format)6th workshop: Added large group sessions; Refined short presentations (maybe consistent format)
• Rapid prototyping?
JLESC: Projects
• Authorized by Executive Committee; involved commitment of resources to travel
• Must be within thematic areas and involve at least two centers
• Must have product defined at the beginning• Most work done ftf in visits• If they work at a distance, mostly via email, filesharing• No consistent project management tools (e.g. JIRA)• Lots of serial collaborators• Report on project on future workshop programs
Profile of JLESC Projects
JLESC Project: Challenges• Many problems given by higher status members (higher
rank, longer tenure in JLESC) to lower status members: extends startup time
• Issue in match between student/postdoc and problem
• Lack of resources central issue
• Leads to incremental problems rather than breaking new ground
• Difficult to bridge application people and computer science people
Applications people “own the code” and trust is slower to develop
JLESC: Visits and Exchanges
• Project centered• Trade personnel—most collaborations are done ftf• The centers have the budgets to do this • Often mentor-mentee relationship• No set format for collaboration; let participants decide• Low-tech collaboration technologies for distance work:
email, skype
Why Does It Work?
• Fit with standard approaches in computer engineering: Intensive interaction to solve a problem that is moderately low structurable (developmental task)
“Software stuff is a contact sport”
• Trust development at project level: Two stages: (1) swift trust through just getting down to work; (2) relational trust through extended interactions
Trust vital to get participants to open up and share ideas without fear of being taken advantage of
• Pressure to produce– For workshops– During time of visits
How Do The Levels Articulate?
• One would like to hope they do
• BUT…plenty of organizations function perfectly well based on lower level persistence, despite a disconnect with higher level steering efforts (competent or incompetent)
• Does the “low overhead” top structure work for the bottom, where the work gets done?
Several interviewees expressed ignorance of what the overall management of JLESC intended (even as they acted in the spirit of the overall enterprise)
Connecting the Two Levels• Collaboration means different things at the two
levels:– At interorganizational level:
• Collaboration means building frameworks to stimulate projects: ideas, session formats (meta)
• Trust is key, but it is trust in other leaders and institutions• Collaboration is effective when metrics—aggregate projects initiated
and completed, etc.—indicate it is– At the project level:
• Collaboration is engaged work together• Trust is trust in your project partners and immediate supervisor• Collaboration is effective when project completed with results
– What makes the connection?
Linking Levels• Institutional Force
– Occupational Culture of Engineers Engineers build things that solve problems.
• Organizational Processes– Downward diffusion of founding culture
• Focus and follow-through; “Made things happen”; “Vision”; Continuity and consistency; Talent
– Shaping due to prevailing view of the task as at best moderately structurable
– Evolution of trust across organizational levels (Curall & Inkpen,2003)• Trust as a decision• Promoted by:
– Habitualization– Organizational fit– Resource commitments
Linking Mechanism
• Downward diffusion process
• Institutional signaling– Institutional messages sometimes– More often signals (unplanned indications)
• Continuous restructuring– Effective alliances go through cycles of(1) formation/negotiation; (2) implementation and operation; (3) evaluation/evolution– Ineffective ones are more static (Curall & Inkpen, 2003)
The Ultimate Dilemma
[New product development] “cannot be a mechanical process of linking steps or coordinating tasks. Rather, effective new product development has to be a creative process of joint learning in which people focus intently on shaping a new product idea and then bring it into existence. In the best of such collaborations, people may not even realize they are collaborating, so that shifting their focus to formal techniques of collaboration may, in fact, reduce their capacity to collaborate.”
Mintzberg, Dougherty, Jorgensen, and Westley, 1996
Questions
• Can this method scale to larger set of centers? What are the limits?
• Modular approach to topics with bottom up project ideas—is there a need for more systematic planning?
• Could greater degree of “system” enable higher levels of participation in projects?
• Are they getting all that they can out of the IVRO?
Acknowledgements
Support from National Science Foundation
• Virtual Organizations as Social Systems Program
• Science Across Virtual Institutes Program
Additional Observations
• An irony: People designing high edge technology are very traditional in how they collaborate
• Status/Power dynamics? Grad students report difficulties in finding collaborators (structural or experiential issue?)
• Interviewees report application area/Computer Science divide is difficult to bridge
• Interviewees also chafe at resource limitations