Upload
michelle-wright
View
219
Download
0
Tags:
Embed Size (px)
Citation preview
Pat LangleyComputer Science and Engineering / Psychology
Arizona State UniversityTempe, Arizona
Challenges and Opportunities inInformatics Research and Education
Thanks to many collaborators for their contributions. This talk reports research funded by NSF, NASA, ONR, and DARPA, which are not responsible for its contents.
The Informatics Revolution
In the 19th and 20th Centuries, the Industrial Revolution brought new tools to aid physical activities.
We are now in the midst the Informatics Revolution, which has brought us new tools for mental activities.
However, to take full advantage of this change, we must:
Educate students in how to use these tools;
Train students to develop improved tools; and
Carry out research on more powerful types of tools.
Drexel University is well positioned to advance the Informatics Revolution on each front.
What is Informatics?
Informatics is a field that designs, develops, and studies artifacts that automate, model, or assist in mental activities like:
Storing and retrieving information content in memory
Encoding and carrying out routine mental activities
Modeling and drawing inferences about situations
Achieving goals via decision making and problem solving
Interacting and exchanging information with others
Formulating creative and innovative responses
Informatics adopts the computational metaphor, but it moves beyond computer science in a number of ways.
Adaptive Interfaces for Personalized Services
Per
sona
lize
d R
adio
Sto
ck A
dvis
or
Tra
vel A
gent
Apa
rtm
ent F
inde
r
A Personalized Travel Agent
A Personalized Travel Agent
This project required developing: a database for airline flights a heuristic search module a graphical user interface machine learning of preferences
We also had to combine them in an effective way.
Ecosystem Dynamics in the Ross SeaEcosystem Dynamics in the Ross Sea
d[phyto,t,1] = 0.307 phyto 0.495 zoo + 0.411 phyto
d[zoo,t,1] = 0.251 zoo + 0.615 0.495 zoo
d[detritus,t,1] = 0.307 phyto + 0.251 zoo + 0.385 0.495 zoo 0.005 detritus
d[nitro,t,1] = 0.098 0.411 phyto + 0.005 detritus
As phytoplankton uptakes nitrogen, its concentration increases and the nitrogen decreases. This continues until the nitrogen is exhausted, which leads to a phytoplankton die off. This produces detritus, which gradually remineralizes to replenish nitrogen. Zooplankton grazes on phytoplankton, which slows the latter’s increase and also produces detritus.
A Process Model for the Ross Sea
model Ross_Sea_Ecosystem
variables: phyto, zoo, nitro, detritusobservables: phyto, nitro
process phyto_loss equations: d[phyto,t,1] = 0.307 phyto
d[detritus,t,1] = 0.307 phyto
process zoo_loss equations: d[zoo,t,1] = 0.251 zoo
d[detritus,t,1] = 0.251 zoo
process zoo_phyto_grazing equations: d[zoo,t,1] = 0.615 0.495 zoo
d[detritus,t,1] = 0.385 0.495 zood[phyto,t,1] = 0.495 zoo
process nitro_uptake equations: d[phyto,t,1] = 0.411 phyto
d[nitro,t,1] = 0.098 0.411 phyto
process nitro_remineralization; equations: d[nitro,t,1] = 0.005 detritus
d[detritus,t,1 ] = 0.005 detritus
We can reorganize these equations as a quantitative process model.
Such a model is equivalent to a standard differential equation model, but it makes explicit assumptions about processes that are involved.
Each process indicates that certain terms in equations must stand or fall together.
process exponential_growth variables: P {population} equations: d[P,t] = [0, 1,] P
process logistic_growth variables: P {population} equations: d[P,t] = [0, 1, ] P (1 P / [0, 1, ])
process constant_inflow variables: I {inorganic_nutrient} equations: d[I,t] = [0, 1, ]
process consumption variables: P1 {population}, P2 {population}, nutrient_P2 equations: d[P1,t] = [0, 1, ] P1 nutrient_P2, d[P2,t] = [0, 1, ] P1 nutrient_P2
process no_saturation variables: P {number}, nutrient_P {number} equations: nutrient_P = P
process saturation variables: P {number}, nutrient_P {number} equations: nutrient_P = P / (P + [0, 1, ])
Inductive Process ModelingInductive Process Modeling
model AquaticEcosystem
variables: nitro, phyto, zoo, nutrient_nitro, nutrient_phytoobservables: nitro, phyto, zoo
process phyto_exponential_growth equations: d[phyto,t] = 0.1 phyto
process zoo_logistic_growth equations: d[zoo,t] = 0.1 zoo / (1 zoo / 1.5)
process phyto_nitro_consumption equations: d[nitro,t] = 1 phyto nutrient_nitro, d[phyto,t] = 1 phyto nutrient_nitro
process phyto_nitro_no_saturation equations: nutrient_nitro = nitro
process zoo_phyto_consumption equations: d[phyto,t] = 1 zoo nutrient_phyto, d[zoo,t] = 1 zoo nutrient_phyto
process zoo_phyto_saturation equations: nutrient_phyto = phyto / (phyto + 0.5)
HeuristicSearch
data: time-series observations
knowledge: generic processes
interpretable process model
phyto, nitro, zoo, nutrient_nitro, nutrient_phyto
variables
Generality of Inductive Process Modeling
acquatic ecosystems protist dynamics
hydrology biochemical kinetics
The Prometheus Modeling Environment
The Prometheus Modeling Environment
This project required that we develop: a new formalism for process models a knowledge base of generic processes constrained search for model structures estimation of nonlinear ODE parameters an appropriate graphical user interface
We also had to combine these elements in an effective way.
More Examples of Informatics Research
Map learning for mobile robot localization and navigation
Visual learning to improve analysis of aerial photographs
Adaptive assistance for crisis-response scheduling
Refining digital road maps using GPS traces from vehicles
Model-driven monitoring of the Space Station power grid
Data-guided revision of a terrestrial ecosystem model
Over the past 15 years, I have also done significant research on:
Many of these efforts have involved collaboration with researchers in fields other than AI and computer science.
Informatics draws on concepts and techniques from computer science, but it differs by emphasizing:
Problem-driven research
address challenge problems, not theoretical issues
System-level innovation
integrate component algorithms, not refine them
User-oriented systems
respond to user needs, not write stand-alone programs
This applied focus makes informatics highly interdisciplinary.
Characterizing Informatics
Application Areas in Informatics
We can organize informatics into broad areas of application:
Health informatics (e.g., order entry systems)
Transportation informatics (e.g., air traffic control)
Military informatics (e.g., command and control systems)
Consumer informatics (e.g., Web recommender systems)
Educational informatics (e.g., intelligent tutoring systems)
Entertainment informatics (e.g., virtual environments/agents)
Science informatics (e.g., in biology, ecology, chemistry)
Because informatics mimics human cognition, it has relevance to all facets of human endeavor.
General Training in Informatics
Both undergraduate and graduate offerings in informatics can teach students to:
Make effective use of existing informatics tools
Understand the principles behind their operation
Gain experience in using them through hands-on projects
Creatively compose them to accomplish complex tasks
Appreciate both their generality and their limitations
Every student would benefit from exposure to informatics tools and their effective use.
Drexel has the opportunity to offer generic informatics courses to its entire student body.
Opportunity
What Is Science Informatics?
collection and storage of scientific data representation and use of scientific models discovery of new scientific knowledge scientific communication and interaction
Science informatics involves the use informatics technology to aid the scientific enterprise.
This broad research area investigates four main topics:
Historically, scientific challenges have often served to motivate informatics research.
Advances in science informatics increase understanding of the scientific process and stimulate new discoveries.
Applications of Science InformaticsA
naly
zing
sky
sur
veys
Clu
ster
ing
gene
exp
ress
ions
Rec
ordi
ng b
rain
act
ivit
y
Ana
lyzi
ng s
atel
lite
imag
es
Applications of Science InformaticsB
uild
ing
onto
logi
esV
isua
lizi
ng s
imul
atio
ns
Bui
ldin
g sc
ient
ific
mod
els
Scie
ntif
ic w
orkf
low
s
The Prometheus Modeling Environment(Bridewell et al., 2007)
An Environment for Systems Biology of Aging
Claims about Science Informatics
Science has always been a computational endeavor; new technology can aid it but not alter its nature.
Information technology is not limited to one facet of science, but cuts across its entire range.
Science informatics rests on general principles that hold across all disciplines.
We can state some interesting hypotheses about the science informatics or e-science movement:
These assumptions have implications for both research and education in science informatics.
Training in Science Informatics
We can train undergraduate and graduate students in science informatics topics like:
the basic structures and processes of science the computational character of science informatics tools that can aid scientific research how these tools operate and principles behind them similarities and differences among disciplines the potential of science informatics and open issues
They should be prepared to use informatics tools in their scientific careers or develop tools for others to use.
Drexel has the opportunity to offer a distinctive minor in science informatics.
Opportunity
A Center for Science Informatics
Drexel would also gain from a center for science informatics that carries out research on computational tools for:
collecting, storing, and managing scientific data creating and simulating scientific models discovering and revising laws and models supporting scientific communities for both general use and for specific fields
Center researchers would collaborate with Drexel scientists in the context of discipline-driven projects.
A few such institutes already exist, but the centers’s breadth would distinguish it and draw international attention.
Opportunity
Candidate Project: Health and the Environment
One promising science informatics project would study the relation between health and environment by:
collecting person-centric data with wearable sensors developing models of how variables are related refining the models to reflect individual differences visualizing both the data and model predictions making results available over the World Wide Web
Such a project would extend science informatics while aiding understanding of environmental effects on health.
In addition, it would support citizen science and aid decisions about environmental and health policy.
Opportunity
Informatics and Virtual Environments
Specify detailed physical settings that reflect characteristics of the real world;
Support visualization and animation of these environments, including interaction with virtual objects; and
Include synthetic characters that operate in the environment and interact with human users.
Another cross-cutting theme in informatics is the growing use of virtual environments that:
Virtual environments have broad applications in entertainment, education, medicine, business, and science.
Within 20 years, most Americans will spend a large fraction of their lives within virtual worlds.
Intelligent Agents and Synthetic Characters
Make inferences about their situations
Carry out activities to achieve their goals
Generate plans that address novel problems
Interact with other agents on joint activities
Synthetic environments pose an ideal setting to drive research on intelligent agents that:
They let us study integrated approaches to embodied cognition in complex but controlled scenarios.
Progress in this area would benefit simulation-based training, interactive entertainment, and other applications.
The ICARUS Cognitive Architecture(Langley, 2006)
Long-TermLong-TermConceptualConceptual
MemoryMemory
Short-TermShort-TermBeliefBelief
MemoryMemory
Short-TermShort-TermGoal MemoryGoal Memory
ConceptualConceptualInferenceInference
SkillSkillExecutionExecution
PerceptionPerception
EnvironmentEnvironment
PerceptualPerceptualBufferBuffer
Problem SolvingProblem SolvingSkill LearningSkill Learning
MotorMotorBufferBuffer
Skill RetrievalSkill Retrievaland Selectionand Selection
Long-TermLong-TermSkill MemorySkill Memory
Synthetic Agents in ICARUSU
rban
Com
bat
Tw
igR
ush
200
8
Mad
RT
S
QuickTime™ and aMicrosoft Video 1 decompressorare needed to see this picture.
Synthetic Agents for Urban Driving
We have developed an urban driving environment using Garage Games’ Torque game engine.
QuickTime™ and a decompressor
are needed to see this picture.
We have created ICARUS agents that combine cognitive and sensori-motor behavior to operate in this complex and dynamic setting.
Synthetic Agents for Urban Driving
We have developed an urban driving environment using Garage Games’ Torque game engine.
QuickTime™ and a decompressor
are needed to see this picture.
We have created ICARUS agents that combine cognitive and sensori-motor behavior to operate in this complex and dynamic setting.
This project required that we develop: a Torque physical driving simulator a specific layout for buildings / streets a formalism for stating ICARUS agents an interpreter for ICARUS programs behaviors for other cars / pedestrians
We also had to integrate these elements.
A Center for Virtual Environments
Urban and natural structures in some environment
Processes and mechanisms that govern their dynamics
Agents that operate in the simulated environment
Knowledge and goals that govern their behavior
Drexel would benefit from a research center that develops new technology for creating and using virtual worlds, including:
The center would develop models at different aggregation levels(buildings, organizations, cities, regions).
Researchers would also develop informatics tools to support the construction, visualization, and simulation of such models.
Opportunity
Candidate Project: A Virtual University
Buildings, streets, and other infrastructure Environmental processes that affect these structures Students, faculty, and staff in the community Knowledge and goals that guide their behavior
One project could involve developing a virtual environment for Drexel University that includes:
Different interest groups could use this model in different ways:
Opportunity
Researchers could test the model against observations Administrators could use the model for decision making Students could visualize the model before attending
Such a virtual presence would make Drexel far more visible, in many senses of the term.
Links to Existing Drexel Activities
College of Information Science and Technology
Department of Computer Science
Biomedical Engineering, Science and Health Systems
Applied Communications and Information Networking Institute
Drexel Engineering Cities Initiative
Human Cognition Enhancement Program
These ideas have clear relations to existing Drexel academic units and research initiatives:
However, their inherent interdisciplinary character makes other connections likely as well.
Some General Activities
Defining challenges and opportunities in an area
Launching a research center in that area
Organizing annual symposia on the topic
Offering tutorials and summer schools in the area
Authoring technical and popular books on the topic
Creating a Web presence and building user communities
Whichever thrusts Drexel pursues in research and education, it can increase its influence and visibility by:
Together, these activities will help make Drexel University a leader in whatever areas it decides to pursue.
Personal Research Themes
Modeling the behavior of complex systems
Developing integrated software frameworks
Combining symbolic with numeric processing
Incorporating insights about human cognition
Building systems that interact with human users
My research trajectory has exhibited some recurring themes:
My future research will continue to follow these principles, independent of the specific problems addressed.
Simon’s Research Heuristics
Be audacious. Tackle challenging problems that others are reluctant to face or even to admit are solvable.
Ignore discipline boundaries. Become familiar with all fields relevant to your research problem and incorporate their ideas.
Use a secret weapon. Take advantage of metaphors and tools that you have mastered but that are not yet widely available.
Balance theory and data. Realize that scientific accounts must respond phenomena but also connect to existing knowledge.
Satisfice. Do not attempt everything at once; idealize your challenging problems enough to make them tractable.
Persevere. Build incrementally on your previous results, extending them to cover ever more phenomena.
The career of Herbert Simon offers guides for scientific research:
I have attempted to follow these principles in my own research.
Concluding Remarks
Training students to use information technology
Developing and understanding new informatics tools
Using problem-driven, system-level, user-centric research
Relating to many socially-relevant application areas
The field of informatics offers challenges and opportunities for:
Drexel University is well situated to take an international lead in this critical area of research and education.
Science informatics and virtual environments, broadly defined, are two themes that hold special promise.
End of Presentation