Title : NSF 97-18--Research in Learning and Intelligent Systems (LIS) Type : Program Guideline NSF Org: CROSS-DIRECTORATE Date : November 25, 1996 File : nsf9718 Research in Learning and Intelligent Systems (LIS) Program Announcement Directorate for Biological Sciences Directorate for Computer and Information Science and Engineering Directorate for Education and Human Resources Directorate for Engineering Directorate for Mathematical and Physical Sciences Directorate for Social, Behavioral and Economic Sciences DEADLINE: February 7, 1997 (Preliminary Proposals) May 15, 1997 (Full Proposals) NATIONAL SCIENCE FOUNDATION NATIONAL SCIENCE FOUNDATION ANNOUNCEMENT RESEARCH IN LEARNING AND INTELLIGENT SYSTEMS Understanding and Enhancing the Ability to Learn and Create The National Science Foundation announces an opportunity for interdisciplinary research in Learning and Intelligent Systems (LIS). Six NSF Directorates-Biological Sciences (BIO), Computer and Information Science and Engineering (CISE), Education and Human Resources (EHR), Engineering (ENG), Mathematical and Physical Sciences (MPS), and Social, Behavioral and Economic Sciences (SBE)-will coordinate and manage the initiative through a special Committee with an appointed Coordinator. The LIS initiative seeks to stimulate interdisciplinary research that will unify experimentally and theoretically derived concepts related to learning and intelligent systems, and that will promote the use and development of information technologies in learning across a wide variety of fields. The long-range goal of this initiative is very broad and has the potential to make significant contributions toward innovative applications. To pursue this goal in a realistic and sustainable fashion, the initiative focuses on fundamental scientific and technological research undertaken in the rigorous and disciplined manner characteristic of NSF-supported endeavors. The initiative ultimately should have a major impact on enhancing and supporting human intellectual and creative potential. Consequently, development of new scientific knowledge on learning and intelligent systems and its creative application to education and to learning technologies are integral parts of this solicitation. BACKGROUND Advanced information and communication technologies will radically transform the way people live, learn, create, and work in the 21st century. The LIS initiative, by integrating technology with research in the many disciplines that contribute to the study of learning and intelligent systems, has the potential to help inform and shape these changes. To do so, the initiative must build on approaches drawn from a wide variety of currently separate but related scientific disciplines and technological advances. The effort to understand the nature of intelligence in general, and the human mind in particular, is one of the most fundamental activities of science. The effort is of old and enduring importance. There are two parallel and compelling reasons why the NSF is framing a new initiative in the general area of learning and intelligent systems at the present time: * There has been a growing convergence of concepts, models, and technologies used in many related disciplines--in the social and behavioral sciences, biology, and in a variety of fields in engineering, computer science, physical sciences, and mathematics--to address issues related to the improvement of information technologies and their application to learning and intelligent systems. For example, there has been a growing use of neural networks, pattern recognition, intelligent control of instrumentation for data collection, visualization, simulation, nonlinear dynamical systems analysis, and probabilistic /statistical learning theory in many of these fields. As another example, the study of synaptic plasticity has provided empirical data to test models for learning, involving the behavioral and cognitive sciences as well as neuroscience and chemistry. Although these models and concepts are used in very different ways in different disciplines, they provide some common foci for research in learning and intelligent systems. * As our understanding of learning, intelligent systems, and information technologies grows, the need to integrate the knowledge generated and to apply it within a broad social context is growing even faster. Research on learning technologies complements and advances interdisciplinary fundamental research in learning and intelligent systems, as it does in other fields. Interdisciplinary research on associated technologies and systems can point to significant breakthroughs in understanding learning and cognition-from empirical research to classroom practice. Integrating research with prototyping in these critical areas promises rapid advances in both theory and application. The initiative challenges interdisciplinary teams of researchers to address two sets of objectives: 1. To identify, investigate, and model the ways natural and artificial systems operate in order to arrive at unifying principles that bear on: * how learning and intelligent behavior occur in human, in other natural systems, and in artificial systems, * the types of learning tasks that are best suited for each, * the kinds of knowledge each characteristically produces or creates, and * the influence of alternative interactive learning environments, social contexts and experiences on them. 2. To enhance the ability to learn and to create by developing a comprehensive set of learning and research tools, and of research methods and technologies that: * integrate biological, behavioral, cognitive, linguistic, social, and educational concepts with interactive, collaborative, and multisensory technologies, and * are accessible to people with varied abilities, knowledge, and expectations. PROGRAM DESCRIPTION While the ultimate goal of this initiative is to understand and enhance people's ability to learn and create, the attainment of this goal requires achievable intermediate goals and strategies. Such strategies include combining the theory, concepts, research tools, and methodologies of two or more disciplines, in order to focus on a tractable component of a larger problem. This initiative seeks to achieve these goals by encouraging interdisciplinary research that has the potential to unify disciplinary knowledge about learning and intelligent systems, and to foster technology research and prototype development to explore and test designs and models that can lead to supportive interactions between natural and artificial systems. LIS-supported research is intended to lead to advances in science and engineering that can foster rapid and radical (as opposed to incremental) growth in the ability to understand and support learning, creativity, and productivity in the natural and artificial systems which are important to a society characterized by significant changes in the complexity of human and information interactions. This solicitation continues the Collaborative Research on Learning Technologies Program (CRLT, NSF 96-80) as an integral part of LIS. A list of FY 1996 CRLT program awards, including planning awards for CRLT Centers, is available on the World Wide Web under http://www.nsf.gov/lis. PRIOR WORK ON A RESEARCH AGENDA Recognizing that research communities are best suited to identify the most critical research challenges and opportunities in the area of learning and intelligent systems, the NSF funded a series of workshops to help plan this initiative. The workshop reports are available electronically as described in Appendix A. Appendix B presents illustrative examples of areas of potential interdisciplinary inquiry . PROJECTS SOLICITED LIS seeks projects that propose: high-risk multi-year research by interdisciplinary teams designed to develop fundamental knowledge that will advance and integrate concepts related to learning and intelligent systems. Projects must go beyond the scope of traditional disciplinary proposals and span the purview of more than one NSF Directorate. experimental prototype systems and technology testbeds that embody theory, test its consequences, and point out factors relating to its eventual efficient application. For the continuing CRLT component of LIS: projects that contribute to the creative integration of basic research in education with basic research in information technology, and that use education at any level (K-12, undergraduate, graduate, life-long learning) as an application domain for basic research in learning and intelligent systems. projects to establish one or more (real or virtual) Centers for Collaborative Research on Learning Technologies (CRLT) to undertake larger collaborative research and development projects, act as a technology transfer mechanism by training new researchers and supporting prototype or model projects, and serve as an evaluation center for learning technology research. It is not required that a proposer have a planning grant from the FY 1996 CRLT competition in order to submit a CRLT Center proposal in FY 1997. AWARDS Research awards are expected to be made for up to $500,000 per year for up to three years. In exceptional cases, awards for up to five years may be considered if the justification and promise are compelling. NSF expects to fund in the order of 20 to 25 research awards, depending on the quality of submissions and on the availability of funds. In addition, NSF expects to fund 3 to 5 learning technology (CRLT) centers, each in the approximate range of $0.5M to $1.5M a year for three to five years. All awards, including CRLT Centers, will be made as grants subject to specified reporting procedures. A total of $19.5M is available for LIS in FY 1997. It is expected that the initiative will continue in FY 1998. REVIEW CRITERIA Eligibility Criteria The development of intellectual integration across disciplines is more than just a constraint that proposals must satisfy. It is a primary goal of this initiative. To be eligible, projects must go beyond the scope of traditional disciplines and span more than one NSF Directorate. Evidence of this interdisciplinary cooperation may take several forms, including the expansion of an established collaboration involving investigators from different disciplines, a credible plan to use NSF support to build such collaborations or links between disciplines, or a proposal from a single investigator who demonstrates the potential for genuine involvement in collaborative multiple disciplinary research. Although established groups of collaborators are eligible for funding, such proposals must provide convincing evidence that LIS funding will lead to advances that have implications above and beyond what would occur without LIS funding. CRITERIA FOR REVIEW All proposals are subject to the guidelines and review criteria described in the NSF publication Grant Proposal Guide (GPG), NSF 95-27. For a description of NSF program activities, refer to: Guide to Programs FY 1996, NSF 95-138. Please note that NSF does not fund clinical research, although it does fund biomedical engineering, as explained in the GPG. Single copies of these publications are available at no cost from the NSF Forms and Publications Unit, (703) 306-1130, or via E-mail (Internet: pubs@nsf.gov). The general review criteria specified in GPG are: (1) research performance competence, (2) intrinsic merit of the research, (3) utility or relevance of the research, and (4) effect of the research on the infrastructure of science and engineering. In addition to the GPG criteria, reviewers will consider the project's degree of innovation and its potential to enhance understanding of learning and creativity. Priority for funding will be given to proposals in which the cross-disciplinary links involve a substantial sharing or blending of paradigms, models, and empirical validation criteria, and to proposals that have the potential to break new ground, mobilize new human resources, or lead to fundamental changes in methodology. Proposals submitted for the LIS initiative will be reviewed based on the following criteria, as appropriate for the type of activity proposed: * The comprehensiveness of the project, including breadth of coverage of the research objectives in relation to the project goals; * The expertise and strength of the research team; its ability to meet the goals of this initiative; and an effective and coordinated management/ leadership plan, evidence that the team can work together; * The likelihood that the results of the research activity will contribute significantly to the intellectual integration of theories, concepts, systems, or methodologies across disciplines; * Proposals that will develop prototypes of tools and technologies also must demonstrate how such technologies will embody intelligent behavior, will augment the human ability to learn and create, or will further research and development of such tools and technologies; * The plans for evaluating, testing, and disseminating the project and its results; * The potential impact of the project on the infrastructure of science, mathematics, engineering and technology (SMET) research and education; for example, by training new researchers to work in interdisciplinary areas, by effectively integrating research and education, or by expanding the breadth of the groups engaged in SMET and in education R&D; * The creation of synergistic collaborations among university researchers or of university collaborations and partnerships with industry, government, schools and school districts, as appropriate; For CRLT Center proposals only: * The knowledge gained from the experience of existing centers; cost-sharing, leveraging of existing and future resources, and the extent of proposed interactions with the outside community; * The effectiveness of the plan for integrating research and education, including but not limited to efforts to engage graduate students and postdoctoral researchers in the project. * The extent to which the Center will develop, test, and/or facilitate the transfer and testing of learning technologies. PROPOSAL SUBMISSION, REVIEW AND AWARD PROCESS PRE-PROPOSALS Pre-proposals are required in all cases and must be received at NSF by FEBRUARY 7, 1997. To help plan for the review process, and in addition to the required preliminary proposal, a short electronic message should be sent to lis@nsf.gov indicating the intent to submit prior to February 7. Pre-proposals should provide sufficient information for reviewers to understand the goals of the project, the quality of the ideas proposed; the strengths of the research team; and the importance of the knowledge to be generated for understanding learning and intelligent systems and for enhancing learning and creativity. The pre-proposal must be single-spaced on letter-size page paper using a 10-point font; and must conform with the following length constraints in the order indicated: Limit to 1 Page: Reference to the LIS announcement (in upper left corner of the page). Title of proposal, and abstract. List of PI and co-PI(s) with department and institutional affiliations, including postal address and telephone numbers, e-mail address and fax numbers. Required: Signatures of the PIs and co-PIs must appear on this page. Institutional signatures are not required by NSF at this stage. Limit to 4-5 Pages: Research and education plans, and a description of interdisciplinary interactions. Preliminary budget summary by budget categories (total amount per year), and project duration. These are expected to be estimates and are not binding for the final proposal. Budget items may include student and post-doctoral support, summer salaries, travel, consultant fees, and equipment when justified by the research or the collaboration, etc. Additional (limit of 1 page/investigator) For each PI and co-PI, provide a brief curriculum vitae, including a list of up to five relevant publications, a list of collaborators during the past five years, and other information deemed relevant . Submissions, referring to this LIS program announcement, including 15 copies of the pre-proposal, must be mailed to: Coordinator, LIS Coordinating Committee NSF, Room 855 4201 Wilson Blvd. Arlington, VA 22230 Pre-proposals meeting the above requirements will be reviewed by internal and external multidisciplinary panels, which will be coordinated and managed across all participating NSF Directorates. Pre-proposals that do not meet the requirements will not be eligible to participate further under this solicitation. It is expected that all principal investigators will be notified of the results of the pre-proposal evaluation by MARCH 1ST, 1997. FULL PROPOSALS NSF will accept full proposals for this initiative by invitation only, based on the results of the pre-proposal eligibility review. Investigators invited to compete should submit a full proposal bearing original signatures and institutional approval. NSF is contemplating the use of Fastlane for full proposals. Detailed instructions will be sent to proposers with the response to the required preliminary proposal, or can be requested from lis@nsf.gov after March 1st., 1997. In addition, one copy of the proposal must be sent to the LIS Coordinator, Room 855 (see above). Proposals must be received at NSF no later than MAY 15, 1997. The submission must follow GPG guidelines, with the following exceptions to the guidelines: On the cover page refer to this announcement; Up to 15 pages, single spaced, are allowed for the project description; Each PI or co-PI may use up to 2 additional pages to describe results of prior NSF support focusing only on those relevant to the proposed project; Each project involving more than one university department, or more than one organization, must provide an additional 1-page description of the project management plan. Subcontracts may be used in multi-institutional proposals. The review process will be conducted by external multidisciplinary panels and will be coordinated across participating NSF Directorates. Inquiries Questions of a general nature regarding this initiative should be addressed via Internet to lis@nsf.gov, or by calling (703) 306-1651. For additional information regarding specific Directorate interests in Learning and Intelligent Systems, you are encouraged to contact several of the following NSF Program Directors: BIOLOGICAL SCIENCES Dr. Christopher Platt, cplatt@nsf.gov (Directorate representative) Dr. Walter Wilczynski. wwilczyn@nsf.gov COMPUTER AND INFORMATION SCIENCE AND ENGINEERING Dr. Larry Reeker, lreeker@nsf.gov (Directorate representative) Dr. Gary Strong, gstrong@nsf.gov EDUCATION AND HUMAN RESOURCES Dr. Nora Sabelli, nsabelli@nsf.gov (Directorate representative; LIS Coordinator) Dr. James Ellis, jellis@nsf.gov ENGINEERING Dr. Kishan Baheti, rbaheti@nsf.gov (Directorate representative) Dr. Paul Werbos, pwerbos@nsf.gov MATHEMATICAL AND PHYSICAL SCIENCES Dr. Deborah Lockhart, dlockhar@nsf.gov (Directorate representative) Dr. Michael Steuerwalt, msteuerw@nsf.gov SOCIAL, BEHAVIORAL, AND ECONOMIC SCIENCES Dr. M. Fernanda Ferreira, mferreir@nsf.gov (Directorate representative) Dr. Steve Breckler, sbreckle@nsf.gov APPENDIX A: Research Reports related to LIS The following LIS-related workshops were sponsored by the NSF. The opinions expressed in these reports are those of workshop participants and do not necessarily represent the views of NSF. The workshop reports are available electronically. The Science and Systems of Learning: Augmenting Our Ability to Learn and Create, September 17-19, 1995 (http://hi-c.eecs.umich.edu/papers/Science_Systems_Learning.pdf). This workshop was convened by a special cross-directorate committee of the NSF, the Committee on Collaborative Research Initiative on Learning and Intelligent Systems. The participants were researchers and educators with backgrounds in the cognitive and social sciences, neuroscience, engineering, computer science, and education. The workshop report provides an overview on research in learning and intelligent systems from a cross-disciplinary perspective that could lead to new ways of learning-- in school, at home, and in the workplace--in the coming age of information. The report also includes illustrative examples of research domains that require collaborative efforts by researchers from several disciplines. Setting a Research Agenda on Educational Technology, September 29-October 1, 1995 (http://www.cc.gatech.edu/gvu/edtech/nsfws/). The objective of this workshop was to focus on ways in which intelligent systems could be useful in all facets of education, including informal and self-directed learning. The workshop report provides examples of interdisciplinary research and systems development related to learning and cognitive functioning that range from empirical research to theory development to classroom practice and that address the application of advanced technologies and new understanding of cognition to the learning process. Information Processing in Biological and Artificial Intelligent Systems, April 8-10, 1996. (http://www-hbp.usc.edu/HBP/presentations/NSF-NISE.Workshop96/). This workshop brought together neuroscientists, physicists, mathematicians, and researchers in AI, control theory, and bioengineering to discuss research opportunities in modeling and understanding of information processing in biological systems, including the brain, for the benefit of studying biological systems and applying such knowledge to artificial intelligent systems. (Publication NSF 97-4) New Horizons in Biosystems Analysis and Control: Analysis, Control and Adaptation of Dynamical Systems in Biology, November 13-15, 1995 (http://robotics.eecs.berkeley.edu/~sastry). The participants in this workshop were from neuroscience, ecosystems, engineering, and computer science. They were brought together to discuss opportunities for collaborative research. The topics included systems identification and analysis tools designed to understand how biological systems interpret sensory signals, control physiological processes, and adaptively monitor and control bioprocesses. The objective was to bring together tools in biology and engineering and identify real-time, nonlinear, stochastic systems capable of learning. Reinforcement Learning, April 12-14, 1996 (http://www.csee.usf.edu/~mahadeva/nsf-workshop/homepage.html). This workshop focused on interdisciplinary research in the area of reinforcement learning. Reinforcement learning is a phenomenon long observed in animals, including humans, that depends on feedback during learning. Reinforcement learning algorithms have been modeled computationally, and the models are being actively studied by an eclectic mix of scientists doing research in machine learning and neural nets, operations research, mathematics, psychology, robotics and engineering. Thirty leading scientists from these fields met to discuss research and applications of reinforcement learning as used for adaptive control and learning in autonomous agents. APPENDIX B: EXAMPLES OF POSSIBLE INTERDISCIPLINARY LIS RESEARCH AREAS The examples provided below illustrate the nature of LIS-related problem domains, the current state of knowledge and understanding of several disciplines, and the fundamental research challenges that require interdisciplinary integration. The examples are provided for illustrative purposes only and are broad in scope. Actual proposals should focus on a tractable component of a larger problem. In addition, the examples are not intended to be restrictive or to target specific areas for this initiative. Additional topic areas include, but are not limited to, mechanisms of attention-focusing, stability of memory versus adaptability, complexity versus building block approach to learning, multiple coexisting parallel models of the world, assessment of learning and creativity strategies, diagnostic tools, intelligent agents in simulated environments, evolutionary constraints and opportunities in learning, the role of emotions in learning. Knowledge representation and multiple levels of abstraction Analyzing complex problems requires people to work with multiple levels of abstraction, simultaneously embedding layers of meaning and representation. Knowing how intelligent systems navigate through such complex representational systems and how to cultivate them are skills that need to be better understood. Engineers have developed systems using multiple models to represent different levels of knowledge abstractions, while some visual systems researchers have used computational engineering approaches to examine how multiple image representations contribute to human perception, and how these perceptual representations are physically represented in the brain. Cognitive science researchers have mapped multiple representations of mental models in humans in several domains, and education researchers have shown real differences between experts' and novices' use of multiple representations during problem solving. Particularly challenging research problems concern how new representational models are constructed, how a user chooses among them at any instant, how existing representations in different domains can be accessed to solve novel problems, and how this skill can be shaped and utilized in an educational environment. Managing information: Selective attention and memory Both in human learning and in artificial intelligent systems, the management of information is becoming a major practical concern. To provide a scientific basis for addressing this concern, a deeper and more multidisciplinary understanding of selective attention and memory management in learning systems is essential. Neuroscientists and cognitive scientists have shown that biological systems have extensive mechanisms for selecting and weighting information. Biologists, biochemists, and psychologists have focused mainly on the internal state of the organism and the attributes of the stimulus. Mathematical scientists and engineers, working in areas such as filtering and statistical decision theory, have developed methods to extract important system parameters from uncertain and noisy measurements. By combining the insights of these various communities, comprehensive selection designs and information management tools could be developed and tested against data from biology and psychology, and evaluated for their engineering utility and rationality. Feedback and learning to optimize/plan How intelligent systems develop strategies, select among alternative plans, and predict the outcome of complex performances is poorly understood, as are the mechanisms by which organisms accomplish these actions. Developing models, technologies and designs that learn to maximize some externally specified measure of performance or utility in complex situations -- situations which require "foresight" or "planning" to optimize performance -- would contribute to our understanding of complex behavior and inform theory development in this domain, as well as assist in the diagnosis and assessment of different learning methods in human and artificial intelligent systems. Engineers and mathematicians have developed learning-based methods for optimization or reinforcement learning that are consistent with flexible neural-like circuitry, but have difficulties in solving complex problems with long time horizons. Computer scientists and psychologists have developed complex, hierarchical planning systems which can handle longer time horizons, but have less flexibility and are less able to handle random disturbances. Motor control research in neuroscience has demonstrated that the brain somehow uses a design intermediate between the two. It would be very useful to generate new artificial learning designs, guided equally by the facts of motor control research and by the challenge of unifying the best features of the competing techniques. Tools and cognitive development Very young children, in what are thought to be early stages of cognitive development, are adept at using diverse technological tools almost as an extension of play, yet it is unclear how tool use affects and enhances learning. Augmenting and facilitating children's cognitive development requires an understanding that goes beyond current theories of developmental stages, and includes developing technological tools that build on children's prior knowledge and that can form a scaffold for the learning of new tasks and concepts. Science education researchers have shown that with appropriate support, young children can design simple experiments and work with abstract concepts previously thought to be beyond their cognitive capacity. Recent advances in computer technology are fostering game-related learning ("plug & play") and are opening novel possibilities for interactive learning environments. The creation of effective learning tools that support and guide individual learning for people of all ages in various settings is a fundamental educational and workplace challenge. Learning and adaptation in hybrid systems Humans, animals, and artificial systems can learn to make discrete choices from a world of perceptually continuous and dynamic variables. However, the relationship between symbolic reasoning and learning and adaptation methods that blend discrete and continuous variables is not well understood. In the cognitive and neural sciences, methods have been developed to analyze and understand dynamic interactions between symbolic (e.g. pattern recognition) and sub-symbolic (e.g., intensity of light) data. In engineering and computer science, concepts from automata theory and dynamical systems have been used to evaluate performance of hybrid systems. Development of analysis and design methods for hybrid systems and testing of data from biology and psychology could lead to better understanding of human and animal motor learning skills and could clarify controversies about the degrees of equivalence underlying learning in humans and animals. Dynamic models of networks A fundamental problem of brain function is how the activity of nerve cells, or networks of nerve cells, relates to behavior. As learning occurs, behavior changes, but the mechanisms for spatiotemporal interactions of neuronal subsystems in the brain are not well understood. Addressing these relationships requires new collaborations. One current approach emphasizes "robustness" of the system, where models are developed without the detail of anatomy and physiology under variable situations, yet the representation produces functionally important behavior. Another current approach emphasizes the functional role that specific properties of the anatomy and physiology might play in studies of particular small networks or of common structure in a variety of networks. Understanding the emergent behavior of subsystems or networks requires synthetic work in areas such as chemistry (structural modeling and host-guest interactions), nonlinear modeling of dynamical systems, neuromorphic chip development for experimental testbeds, and analysis of how coupled oscillators in biological systems learn to produce coordinated behavior. Distributed intelligence: Pattern recognition and categorization Groups and organizations as well as artificial systems use distributed intelligence to perform their work. In these systems, rules are not explicitly encoded; no one part possesses much knowledge but the parts working together learn and produce intelligent behavior. How these systems -- systems that address problems of societal importance such as categorizing weather hazards and identifying the properties of economic markets -- perform the often daunting tasks of pattern recognition and categorization in the face of uncertainty and contaminated data needs further understanding. Researchers in such diverse fields as the physical sciences, economics, and meteorology have developed intelligent characterization tools and innovative systems for performing pattern recognition and categorization tasks, but may be unaware that their work contributes towards the understanding of learning and intelligence. Some systems take the form of neural nets and nonlinear dynamic models. Others involve use of techniques incorporating mathematical analysis of images within local elemental units of the data space, synthesis of these disparate components using fuzzy logic methods, the extraction of desired global features, and intelligent control of instrumentation. The research challenge is to bring together scientists from fields not typically associated with the study of learning and intelligence with scientists from the cognitive, neural, and computer sciences and engineering, so that fundamental properties of systems that learn may be identified and practical problems of creating supporting environments may be addressed with greater success. Learning-based Management of Massive Data Sets Modern data acquisition technologies can produce overwhelmingly large data sets which challenge human interpretation. In the geosciences, for example, this commonly takes the form of many physical parameters and dependent variables referenced to a very large number of geographic coordinates. In high energy physics, devices process extremely high particle densities, a wide range of particle momenta, and extremely high event rates. These and related problems in other physical sciences require sophisticated high-speed data acquisition electronics both to process the data rapidly and to find relevant, often rare, events hidden in a much larger set of events. All across the observational sciences there is a pressing need to manage such data so as to maximize what can actually be learned from it. New technologies are being developed to meet this challenge by combining digital acquisition and processing technologies with the development of intelligent systems for assimilation, analysis, and comprehension of very complex data. New research could try to improve the effectiveness of learning-based systems in these applications by developing and exploiting a better understanding of the cognitive processes involved in perception, pattern recognition, computation, multivariate analysis, and spatial analysis. Likewise, collaborative research teams could develop new learning-based technologies to perform these kinds of tasks in a more intelligent and flexible manner. ADDITIONAL INFORMATION The Foundation provides awards for research in the sciences and engineering. The awardee is wholly responsible for the conduct of such research and preparation of the results for publication. The Foundation, therefore, does not assume responsibility for the research findings or their interpretation. The Foundation welcomes proposals from all qualified scientists and engineers and strongly encourages women, minorities, and persons with disabilities to compete fully in any of the research related programs described here. In accordance with federal statutes, regulations, and NSF policies, no person on grounds of race, color, age, sex, national origin, or disability shall be excluded from participation in, be denied the benefits of, or be subject to discrimination under any program or activity receiving financial assistance from the National Science Foundation. Facilitation Awards for Scientists and Engineers with Disabilities (FASED) provide funding for special assistance or equipment to enable persons with disabilities (investigators and other staff, including student research assistants) to work on NSF projects. See the program announcement or contact the program coordinator at (703) 306-1636. Privacy Act and Public Burden. The information requested on proposal forms is solicited under the authority of the National Science Foundation Act of 1950, as amended. It will be used in connection with the selection of qualified proposals and may be disclosed to qualified reviewers and staff assistants as part of the review process; to applicant institutions/grantees; to provide or obtain data regarding the application review process, award decisions, or the administration of awards; to government contractors, experts, volunteers, and researchers as necessary to complete assigned work; and to other government agencies in order to coordinate programs. See Systems of Records, NSF 50, Principal Investigators/Proposal File and Associated Records, and NSF-51, 60 Federal Register 4449 (January 23, 1995). Reviewer/Proposal File and Associated Records, 59 Federal Register 8031 (February 17, 1994). Submission of the information is voluntary. Failure to provide full and complete information, however, may reduce the possibility of your receiving an award. Public reporting burden for this collection of information is estimated to average 120 hours per response, including the time for reviewing instructions. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Herman G. Fleming, Reports Clearance Officer, Contracts, Policy, and Oversight, National Science Foundation, 4201 Wilson Boulevard, Arlington, VA 22230. The National Science Foundation has TDD (Telephonic Device for the Deaf) capability, which enables individuals with hearing impairment to communicate with the Foundation about NSF programs, employment, or general information. To access NSF TDD, dial (703) 306-0090; for FIRS, 1-800-877-8339. Programs described in this publication are within the following categories of the Catalog of Federal Domestic Assistance (CFDA): 47.041, 47.049, 47.070, 47.074, 47.075, 47.076 OMB: 3145-0058 NSF 97-18 P.T.: 04, 18, 35, 36, 38 [Replaces NSF 96-80] KW.: 0414007, 0503000, 0607004, 0607070, 0706000, 0710015, 404000, 410000, 414000, 1002030, 1004000, 1010000, 1016000