LINK: Learning and Innovation
in Networks of Knowledge
Enabling the Stretch Organization
David E. Goldberg, Michael Welge, & Tim Wentling
University of Illinois at Urbana-Champaign
Urbana, IL 61801
{deg, mwelge,wentling}@uiuc.edu
University of Illinois at Urbana-Champaign National Center for Supercomputing Applications

Motivation
• Traditional organizations designed to
perform routine tasks:
– Design
– Manufacture
– Sales/Service
• Works well at core business under normal
times.
• Traditional organization may have difficulty
responding to
– Crisis
– Opportunity
• Can the organization stretch using HR & IT Crisis: wei-ji
• To handle the crisis and/or exploit the
opportunity.
National Center for Supercomputing Applications

Foundation: Collaboration & Experience
• Collaboration of three laboratories:
– Automated Learning Group (ALG), NCSA
– Knowledge and Learning Systems Group (KLSG)
– Illinois Genetic Algorithms Lab (IlliGAL)
• Experience set:
– Knowledge Management
– Learning Sciences
– Data and Text Mining
– Collaborative IT Systems
– Genetic Algorithm Research
– Innovation Research
National Center for Supercomputing Applications

The Stretch Org: Human-Centered IT
• Human-centered IT can help organizations stretch:
– HCIT can aid collaboration
– HCIT can capture tacit and explicit knowledge
– HCIT can aid innovation across disciplines and across
contexts
– HCIT can tie stakeholders at the periphery to the core.
– HCIT can use data to learn and calibrate enterprise-
central models.
– HCIT can gather other computational tools and
procedures to present array of high quality possibilities.
National Center for Supercomputing Applications

LINK Goals
• Create integrated HCIT environment for the
stretch organization.
• Gather
– collaborative tools
– knowledge communities and stores
– learning/mining capability
– augmented innovation
– models management environment.
• Place humans, computers, and computational tools
at focus of any crisis or opportunity.
National Center for Supercomputing Applications

LINK Builds on D2K, DISCUS, & KC
• Pieces before 2003.
• Integrating impulse: Collaborative systems for innovation
support.
• NCSA ALG & KLSG, and IlliGAL
• Industrial partner: Hakuhodo
• Japanese academic partners: Osawa-Lab shortly thereafter
& Nao Matsumura somewhat later.
• Navy, industrial, and AF funding.
• Has resulted in
– Patent issued, and 5 patent pending on core technology.
– Pilot and actual commercial trials.
– 3 entrepreneurial spinoffs.
National Center for Supercomputing Applications

Roadmap
• The innovation intuition.
• 3+2 challenges of online innovation.
• Putting the Zoom in Zoom-Zoom-Zoom.
• LINK in Combat: Semantics, adaptation &
influence in networked teams.
• Modeling user preference functions.
• Looking for LINK Consortium partners.
National Center for Supercomputing Applications

The Innovation Intuition
• Two reasons:
1. The innovation intuition
2. The human-computer quad chart
• The innovation intuition
(Goldberg, 1983, 1989, 2002):
1. Selection + Mutation = Kaizen
2. Selection + Crossover =
Crossfertilizing innovation.
National Center for Supercomputing Applications

Selection + Recombination = Innovation
• Combine notions to form ideas..
• “It takes two to invent anything. The
one makes up combinations; the other
chooses, recognizes what he wishes
and what is important to him in the
mass of the things which the former
has imparted to him.” P. Valéry
National Center for Supercomputing Applications

Human-Computer Collaboration
computational
Innovation agent
Standard Interactive
Genetic Algorithms Genetic Algorithms
human
Computer Aided Human Based
Design (CAD) Genetic Algorithms
computational human
Selection agent
(Kosorukoff & Goldberg, 2002)
National Center for Supercomputing Applications

3+2 Challenges of Online Innovation
• Initial three design challenges:
– Online superficiality.
– Variation in stakeholder articulateness.
– Quantity & heterogeneity of data and analytics sources.
• Two more:
– Inadequacy of fixed structure.
– Importance of the social network of influence.
National Center for Supercomputing Applications

Online Superficiality
• Online communication
is superficial.
• Tendency toward
write-only world.
• Many distractions.
• Difficult to reflect.
Primary solution:
Enhance online
reflection &
innovation through
semantic visualization
National Center for Supercomputing Applications

Variation in Articulateness
• Different stakeholders
can express themselves Global Stakeholders
Population (GSP)
differently. IEC
• 2 types: Collaborative Validation
Community (CVC)
– Explainers Ke
GA yG
HB ra
ph
s
– Evaluators Core Coherent
Innovation Team
(CCIT)
T2 K
D2
K
Primary solution:
Use full collaboration
at center and
interactive GA at the
boundaries.
National Center for Supercomputing Applications

Interactive GAs
• Interactive GA replaces objective
function with subjective function
from a human.
• Famous example: Faceprints
(Caldwell & Johnston, 1991).
• Criminal face reconstruction.
• Used from art to marketing
(www.affinnova.com).
Figure : Actual photo of simulated criminal
(above). Evolved image from witness using
Faceprints (right). (Caldwell & Johnston,
1991)
National Center for Supercomputing Applications

Quantity & Heterogeneity of Data
• Applications need support by lots of data.
• Also need problem-specific analytics.
• Want these at innovation fingertips.
Primary solution:
Build in capable
data-text miner (D2K)
and provide flexible
web services interface
to problem-specific
tools.
National Center for Supercomputing Applications

Inadequacy of Fixed Structure
• Two versions tried two extremes of organization.
• Amorphous discussion and fixed brainstorming
protocol.
• Neither sufficiently flexible for applications
Primary solution:
Build increasingly flexible administrative controls
for space and time topology.
National Center for Supercomputing Applications

Mining the Network of Influence
• Were focused on semantics to exclusion of the social network.
• Thought that network visualization might be as powerful for reflection as social.
Primary solution:
Incorporate influence diffusion methods.
• On an ongoing communication we can measure
– Key messages
– Propagation of terms/concepts
– Measure influence of people based on propagation
• Visualize:
– The trace of influential terms
– The social network of influence (qualified subset of the SN of communication)
National Center for Supercomputing Applications

Quantifying Influence on a Real Focus Group
Exclude high frequent terms
National Center for Supercomputing Applications

Navigating the Influence
National Center for Supercomputing Applications

Visualizing the Network of Influence
• Help identify discussion roles + identify behavior patterns
National Center for Supercomputing Applications

Web Services Architecture
DISCUS Applications DISCUS Integrated Web Interface (DIWI)
DISCUS Service Abstraction Layer
Functional Scenario Definition
DISCUS Workflow
Scenario Engine
DISCUS Abstraction Layer
Collaboration Integration Layer iGA Chance Discovery HBGA Models Visualization
On-line archiving Topic tracking/identification Core I2K T2K Theme Weaver
Conference Collaboration Knowledge
D2K Engine Webcrawlers
Technology Infrastructure Sharing
National Center for Supercomputing Applications

Piloting Near-Commercial Prototypes
• Need user
ready testbed.
• Use standard
discourse
elements.
• Academic++
code
(commerce-).
National Center for Supercomputing Applications

Keygraph & HBGA Box
National Center for Supercomputing Applications

Interoperability of Tools
National Center for Supercomputing Applications

Putting the Zoom in Zoom-Zoom-Zoom
• Work w/ Hakuhodo, no. 2 Japanese advertising firm.
• DISCUS in use for design of concept car by Mazda.
• Innovation in marketing v. warfighting: Changing
consumer vs. changing adversary.
National Center for Supercomputing Applications

LINK in Combat Coordination: SAINT
• SAINT: Semantics, Adaptation, and Influence in
Networked Teams
• Funded LINK extension for enhancement of
command communication and innovation to US
Air Force.
• Key elements of the proposal:
– Tighter integration of influence and semantics.
– Seamless integration of humans and computers.
• Seeking integration into Air Operations Centers
(AOCs).
National Center for Supercomputing Applications

Toward Seamless H-C Creativity
Building block formation and mapping. Mine key
terms and relations and form building blocks or
conceptual clusters.
Chromosome formation and interaction. Develop
modified vector approaches moving toward Bayesian
models of term co-existence.
Building block social evaluation. Use IDM process for
social evaluation of building blocks.
Conceptual chromosome fitness evaluation. Build
fast, accurate models of human preference to be used
in lieu of H evaluation.
Idea generation. Combine subsystems for effective HC
creative idea generation.
National Center for Supercomputing Applications

User Subjective Functions
Evaluation
Partial order
update
Sampling Sampling
fs(x)
Optimizer
Synthetic fitness
National Center for Supercomputing Applications

Pilot Experiment
National Center for Supercomputing Applications

Building Block on Communication Channels
• Example conducted on one email account
• 450 emails approx.
• Low traffic email box
• No spam
National Center for Supercomputing Applications

Social Network Analysis
• Email box content extraction and processing
• Following the DISCUS experience we heavily used
visualizations
• Visualization of data and analysis results
• A first step involve identifying key players
– Social network visualization
– Degree ranking distribution analysis visualization
– Betweenness/centrality analysis visualization
– HITS analysis visualization
• IDM vs SN analysis provides a way analysis corporate
versus community communication patterns
National Center for Supercomputing Applications

Social Network Visualization
National Center for Supercomputing Applications

Degree Ranking Distribution (Input)
National Center for Supercomputing Applications

Degree Ranking Distribution (Output)
National Center for Supercomputing Applications

Betweenness/Centrality
National Center for Supercomputing Applications

Content Analysis
• Email box processed content analysis
• Visualizations
– Key self/non-self discriminant terms.
• For a given user, we build a model that distinguishing self
from non-self email.
• It works remarkably well.
• Terms can be ranked according to their weight.
• Then all terms are rendered and connected based on weight
and grouped by users (useful to identify overlapping and
differences among user’s topics)
– Communication topics
• Identification
• Transition detection
National Center for Supercomputing Applications

Self/Nonself Discriminating Terms
National Center for Supercomputing Applications

Term Dynamics
National Center for Supercomputing Applications

Term Dynamics
National Center for Supercomputing Applications

Detecting Topic Transition in a Real DISCUSsion
navigation
gps
service
mp3
music
internet
talk
player
bluetooth
call
batteri
function
send
screen
voicemail
smaller
avoid
simply
feel
keyboard
answer
feature
function
technology
depend
problem
life
computer
National Center for Supercomputing Applications

Looking for LINKs: Consortium Partners
• You want LINK, LINK wants you.
• Discussions underway with major
insurance company to use LINK to
stretch organization.
• Looking for industrial partners to
move beyond E-mail to innovation
support.
• Ground-floor opportunity: Initial
partners will set agenda and
determine pilot studies.
LINK
• Those interested should contact
Brian Kucic for more information:
bkucic@ncsa.uiuc.edu
National Center for Supercomputing Applications

Summary & Conclusions
• LINK: Learning and Innovation in Networks of
Knowledge.
• Using HCIT to help organizations stretch:
organizational innovation during times of crisis or
opportunity.
• Online mining of meaning and influence together
with creativity support mechanisms.
• LINK consortium forming: First companies will
get an innovative jump on the competition.
• Find out more about the LINK Consortium, today.
National Center for Supercomputing Applications

More Information
• KLSG:: http://learning.ncsa.uiuc.edu/
• IlliGAL: http://www-illigal.ge.uiuc.edu/
• ALG: http://alg.ncsa.uiuc.edu/
• DISCUS: http://www-discus.ge.uiuc.edu/
• Recent book: Goldberg, D. E. (2002). The
Design of Innovation. Boston, MA: Kluwer
Academic. http://www-doi.ge.uiuc.edu/
• IlliGAL blog: www.illigal.blogspot.com
• Consortium Info:
jmnolan@ncsa.uiuc.edu
National Center for Supercomputing Applications