Large disasters like Hurricane Katrina offer familiar
Responders need to gather information from diverse
and unexpected sources.
Responders also need to use the information
There is limited information access in the disaster zone.
But many of the important decisions are made
An agile, ad hoc response is needed.
Modern technology must support this.
There is a rush to organize:
Ephemeral groups (Farnham)
Hastily Formed Networks (Denning)
A trade-off emerges between:
Creative Response to:
Thousands of people spontaneously collaborate over the
There are several aspects to collaboration if enormous
virtual teams are to be effective.
Participants must to come to agreement on:
The problem definition
They must capture information and know how to pass
it to those who need it.
They must make decisions by forming a consensus
among massive numbers of participants
Teammates establish common ground by
combining their individual mental models of
the problem into a team model.
Transmission of Information to the Appropriate Expert
This collaboration tool will go beyond existing social
The basic functional unit will consist of two interfaces:
The data entry interface will allow easy
entry, categorization, and visualization of large amounts of
The interaction interface will support the formation of ad-hoc
teams and the engineering of collaboration protocols for
negotiation of coordinated action.
These basic units will be coordinated with a
third, agent-augmented mixed-initiative interface.
In this way, we hope to break any large problem into
small pieces and solve it in a coordinated manner.
This approach allows the combination of unstructured
chat with structured knowledge-building.
It splits a team of unlimited size into unlimited sub-
teams of limited size.
The communication and processing power of the
Internet can then be used to coordinate both intra-
group and inter-group interactions of these sub-teams.
Users must have the ability to gain access to the MCT.
Users must be interested in joining a team and in
helping each other.
Users must be able to learn the interface quickly and
under stressful conditions.
Users must understand both the interface and the
subject matter well enough to develop and negotiate
data models and action plans.
Initial use cases
Paper Prototype -> Focus Groups -> Specs
Working Prototype of the Sub-Team Interface
Phase 1 Prototype – Initial testing of the MCT
Effects on team-building and decision-making
23 participants -> 4 Test Teams, 4 Control Teams
Test team had full interface, control team only chat
Used pre/post surveys, analyzed chat/artifacts
Phase 2 Prototype – Redesign and Follow-Up Testing
10 participants for requirements gathering using previous
10 participants -> 5 using previous interface, 5 using new one
Tested five basic data entry and manipulation tasks
Motivating Goal for
Local Emergency District Fire Determination of
Responders Superintendent Priorities
Volunteer Labor Firefighters’ Union Resource Coordination
Non-Profit Aid Resource Coordination
Red Cross Coordinator
National Guard Response Activity
Federal Emergency Jurisdiction
Concerned Common Resource Donation
Volunteer Workers Social Worker Resource Donation
Volunteer Experts Computer Expert Technology Donation
Rescue of Family
Affected Individuals Relative
Victims and Families
Types of Interaction
Person to person
Group to group
Human to agent
ID Interaction ID Interaction
1 Find Site 10 Develop Mental Models
2 Use Site 11 Negotiate Group Models
3 Find Area of Interest 12 Vote
4 Participate 13 Take Turns
5 Converse 14 Exchange Information and Resources
6 Create Team 15 Form Teams of Agents
7 Join Team 16 Agent-Mediated Playoffs
8 Leave Team 17 Inter-Group Negotiation
9 Disband Team 18 Provide Help
Users wanted flexible categories.
Temporal versus logical organizations
Users wanted capable data handling.
Cut and paste
Users wanted flexible interaction sequencing.
Emergent leadership helped and should be supported.
The merging of models increased data organization.
Rapid input and output of data
Abstraction and Type Development
Move from a set of everything…
To many sets of specific types of things
Define the relationships between the sets
Requires cut and paste of entire structures
Initial collaboration design in prototype:
Each individual creates a data model.
Individuals look at each others models.
Individuals take turns building the team model.
Team elects a leader.
Leader develops action plan.
Possible alternative design using thinkLets
Ultimate Goal: The team dynamically designs the
A thinkLet is a facilitation intervention that creates a
predictable, repeatable pattern of collaboration among
people working together toward a goal.
It includes everything a designer needs to reproduce
this pattern of collaboration.
It is the smallest unit of collaborative activity that the
process designer can manipulate.
ThinkLet name – descriptive and/or metaphorical
Choose this thinkLet… – list appropriate uses
Do not choose this thinkLet… – list inappropriate uses
Overview – give a brief narrative description
Inputs – enumerate what is needed to start
Outputs – enumerate what deliverables will result
Setup – describe the necessary preparations
Steps – describe each step of the procedure
Insights – discuss how and why it works, tips, pitfalls
Success stories – describe examples of successful field use
What’s in a name – explain how the thinkLet got its name
User testing to date has used static scenarios.
But lives hang in the balance in a real disaster.
The MCT must be tested in a safe environment that effectively
mimics real-world disaster conditions.
With the expansion of our research team we have gained access
to such a test bed.
The NeoCITIES disaster simulator is
A computer-based scaled-world simulation
Designed to mimic the situation assessment and resource allocation
tasks of distributed emergency crisis-management teams.
The group activity consists of distributed individuals:
Jointly gathering information about emergency events,
Allocating the appropriate type and quantity of resources to
address these events
Detecting emerging threats and patterns of activity from an
This experimental approach provides a holistic
assessment of distributed cognition by:
Providing real-time challenges
Allowing the tool to be used in addressing these challenges
Supporting intra and inter team communication measures
The autonomous software agents use data-mining
techniques to compare the models of all the sub-teams.
Small-worlds networking principles are used to link
the agents to each other.
Collaboration among the agents then leads to
suggestions for collaboration among the sub-teams.
Representatives from each sub-team are sent to a
Information from the playoff team then feeds back to
PhD Candidate Informatics, Indiana University, Indianapolis, IN
Executive Associate Dean, IU School of Informatics
Director, Media Informatics and Human-Computer Interaction
Ph.D. Purdue University, West Lafayette, IN (2005)
Associate Professor, IU School of Informatics
Ph.D. Computer Science, University of Cambridge, UK (1996)
M.S. Candidate HCI, Indiana University, Indianapolis, IN
Assistant Professor, IU School of Informatics
Ph.D. Information Sciences and Technology
The Pennsylvania State University (2008)
Gert-Jan de Vreede,
Kayser Distinguished Professor and Director, Center for
Collaboration Science, University of Nebraska at Omaha
PhD Organizational Change, Delft University of Technology