Gautam Biswas - US grants

9873520
Bransford

This project combines insights from recent work in computer science, psychology and education to create and study "teachable agent" (TA) environments in mathematics and science that are motivating to students, intuitive to teachers and parents, and lead to high degrees of student learning. The hallmark of these environments is that students learn by instructing "teachable agents" who then venture forth in simulation-based exploratory environments and attempt to solve problems that require knowledge relevant to the disciplines of mathematics or science. If the agents have been taught properly they solve the problems; otherwise they need to be educated further. The simulation-based environments are carefully designed to focus attention on important concepts in science and mathematics, and to make explicit the errors that occur during problem solving. Students "scout" the problem solving requirements of various environments before attempting to teach their agents. Additional help and coaching agents are available to point students in the right direction when they make errors or produce sub-optimal solutions. The focus of the TA environments is on learning standards-based content in science and mathematics, not on learning to program.

One key issue to be studied is how student learning is affected by opportunities to teach agents to prepare for particular sets of challenges. Also to be studied is how learning is influenced by the design of systems that vary in the degree to which they let students (a) "scout" to find problems that arise in the agents' environments; (b) teach the agents with different representations and techniques; (c) measure the successfulness of their teaching by placing their agents in mini-assessment environments prior to engaging them in full-blown "challenge environments"; (d) receive different degrees and forms of feedback when their agents encounter difficulties; and (e) educate the personality, as well as the knowledge variables, relevant to learning, problem solving and collaboration.

The project requires contributions from, and has important implications for, at least three disciplines: computer science, psychology, and education. The project has the potential to create new forms of assessment, and to transform popular video technologies into environments that help students learn important content. The quality and impact of the project will be enhanced through its association with the NSF-funded Center for Innovative Learning Technologies (CILT), whose mission is to foster collaboration among members of the education and technology community.

The goal of this project is to explore an innovative model of statistics instruction and assessment that prepares students to learn more effectively from many styles of instruction, including traditional lectures. Many children have difficulty learning to use and understand mathematical representations. One of the reasons for this difficulty is that children often do not understand the work a given representation has been designed to accomplish. For example, they tend to view symbolic formulas as rules that make life harder, not easier. This project will study classroom environments that help middle-school students appreciate the work that statistical representations need to perform, and in this way, it will prepare them to learn about those representations. A central feature of these environments is that students invent their own procedures and representations for solving problems about quantitative situations. For example, if students are learning about variability, very simple situations might include the sets of numbers {2 4 6 8} and {4 5 5 6}. The students are reminded that there is a procedure for determining a single value that captures what is the same about the sets; namely, the average. The students' task is to build a method for determining a single number for each set that can characterize what is different, in this case, the variability. After they create their own methods (often a range formula), they receive new sets that draw their attention to other properties of distributions (e.g., the middle numbers -- {0 2 4 6 8} vs. {0 0 0 0 8}). After several cycles of inventing, assessing, and revising, the students are prepared to learn about and use conventional approaches to variability (e.g., standard deviation). The goal is to help students notice important quantitative properties, and to prepare them to learn how conventional representational solutions elegantly capture those properties.

Biswas, Gautam
CCR-0208799

Embedded systems, which include software components integrated into physical processes, are per-vading all aspects of our daily lives from home appliances to safety-critical systems, such as aircraft and nuclear plants. The widespread use necessitates the development of new engineering tech-niques that can ensure their timely and assured development, accurate monitoring during operation, and robust control, to ensure safety and reliability. This project develops an integrated model-based approach to embedded system development that includes the plant, its environment, and the embed-ded computing system. The models provide a common framework for design and run-time analyses of system stability, liveliness, safety, security, and real-time supervisory control. Models can also form the basis for generating the hardware and software components of the embedded systems, and de-fining their run-time configurations. This generative aspect of modeling is a very relevant and distin-guishing property of the model-based development process. Further, the ability to analyze system behavior at run-time forms the basis for methodologies designed to accommodate deviations caused by disturbances and unexpected changes in the environment. The goal is to ensure that the system and its surroundings are not harmed when aberrant situations occur.

The project will develop technologies for run-time analysis of embedded systems that alleviate some of the complexities of modeling and analysis of systems with large mode spaces. Specifically, effec-tive methodologies that address run-time dynamic analysis issues are addressed. This includes three primary tasks:
Developing a new concept called the dynamic hybrid automaton (DHA) for embedded sys-tems models with large mode spaces. A DHA is simply a hybrid automaton, which can be constructed incrementally, on-the-fly, at run-time, as system behavior evolves. It is based on formal compositional modeling techniques in the hybrid automata framework that ensure model construction grows linearly (as opposed to exponentially) as a function of the number of switching elements in the hybrid model.
Tracking system behavior using hybrid observers developed from the hybrid automata mod-els. This involves techniques for updating the models of the observer on-line when a mode change is detected while tracking the plant behavior. Research challenges focus on model and tracking procedures that minimize mistracking at mode transition boundaries, and devel-oping code generation systems that allow for incremental recalculation of the observer mod-els while satisfying hard time bounds on the generation process, and
Synthesizing supervisory controllers on-line in response to mode changes, some of which may be attributed to disturbances and unexpected changes in the environment. A new con-cept, the Active Controller Model (ACM), is proposed. The ACM is a dynamic data structure that explicitly represents the currently active supervisory controller (SVC), together with its generator and actuator. The SVC can be implemented as a generic procedure that uses the ACM as its "knowledge base" to compute what control actions to take. When the plant model changes, the ACM is updated to address the new situation. This will involve a number of in-novative research tasks, such as developing an expressive language to describe control ob-jectives, definition and incremental update procedures for the ACM models, and "anytime" re-source-bound algorithms for synthesizing supervisory controller code on-line. Robust super-visory controllers will extend the concept of adaptive control into the hybrid-systems domain, and adjust to configuration changes in the plant and environment.

The success of all three components of this project is very heavily dependent on handling computa-tional complexity issues in incremental model generation, code generation for the hybrid observer, and on-line supervisory controller synthesis based on desired objectives for the plant. Therefore, complexity studies of the synthesis and code generation algorithms is an important component of the project. The goals are ambitious, but the success of these methods will offer new flexibility in embed-ded applications while addressing issues of reliability and safety during run-time operation.

It is a common intuition that one of the best ways to learn something is to attempt to teach it to others. The research project is testing this intuition and its usefulness for educational practice. Central to the work are web-based "teachable agents" that students teach as a way to learn themselves and that permit studies of the active ingredients of learning by teaching. The project will use the agents to help students learn science and mathematics content and help teachers learn content-specific pedagogy. The intellectual merit of the work is that the agents provide a method for developing and testing a learning theory that helps explain the motivational and cognitive benefits of learning by teaching. By manipulating a teachable agent's available features and use (something that is hard to do with actual pupils), it is possible to experimentally isolate different components of the learning by teaching interaction. A primary measure used to evaluate whether students learn by teaching includes assessments of students' abilities to subsequently learn from resources once they have completed their teaching activities. The broader impacts include a new way of helping teachers understand the learning of their students, and the development of easily distributed, software that embodies new learning principles so that others in schools and industry can learn from those artifacts and create new ones of their own.

CNS-0452067 Gautam Biswas Title: SGER: Modeling, Analysis, and Diagnosis for Safety of Distributed Hybrid Systems

Distributed embedded systems are pervading all aspects of our daily lives, from home appliances to safety and mission critical systems, such as automobiles, aircraft, manufacturing processes, nuclear plants, and military systems. These engineered systems consist of multiple subsystems with software components tightly integrated into physical processes. Complete analysis of their behaviors at design time is computationally infeasible. Diagnosis of fault behavior and its potential propagation through highly-coupled complex systems lacks a comprehensive scientific foundation. Commonly used methods for fault and failure mode criticality analysis do not address the detailed interactions of these systems and do not scale. A key observation is that the coupled subsystems interact at the physical level through energy-related interactions and at the logical level, by information exchange typically facilitated by a communication fabric such as a local area network LAN or control network.

The objectives and primary thrust areas of this exploratory research effort are two-fold.
(i) Advance the scientific understanding of modeling and behavior analysis of complex, embedded systems that are made up of distributed, interacting subsystems. This will require developing modeling methodologies and languages that integrate heterogeneous paradigms, developing formal models and mechanisms to handle subsystem interactions, and parameterizing the models to facilitate fault analysis.
(ii) Develop practical technologies for safety analysis and online diagnosis of complex distributed systems. The significant challenge in building a systematic framework for distributed diagnosis is to specify the coupling between subsystems in a way that fault interactions are captured in sufficient detail and integrated into the hybrid modeling methodology. It is also important to combine the results of the subsystem diagnostic components in a computationally efficient manner.

The science and technology developed in this project will inform designers in how to build more effective, reliable, and verifiable systems. For rapid dissemination through graduating engineers, research results will also be introduced in undergraduate and graduate engineering classes and laboratories at Vanderbilt University. Besides regular publications, efforts will be made to run a focused workshop in this area.

Distributed embedded systems that consist of multiple interacting subsystems with tightly integrated software components are pervading all aspects of our daily lives. Online fault detection and isolation (FDI) is the key to the safe and reliable operation of these safety critical systems that include automobiles, aircraft, hospital equipment, manufacturing processes, military systems, and nuclear plants. Contrary to its importance, FDI for the safe operation of large, distributed embedded systems is not a solved problem. As a result, in present day systems, almost all fault management and remedial tasks are left to human operators, who often face information overload and stringent time constraints when operating these systems in mission critical operations.
The objective of this project is to develop systematic, scalable, robust, online model-based FDI schemes for distributed embedded systems. The novelty of the research centers on (i) hierarchical abstraction schemes for managing the complexity of the FDI task and enabling the design and development of online model-based FDI algorithms that are provably robust and reliable, (ii) a unified framework for diagnosis of multiple types of faults that occur in the physical and the computational parts of embedded systems as well as faults with different fault profiles (abrupt and incipient faults), and (iii) the development of a tool suite for distributed embedded systems for online FDI. Experimental test-beds are used to demonstrate and verify the effectiveness of the developed methods. The impact of the project lies on providing guarantees for reliable safe operation of complex, distributed safety-critical systems.

An important goal of formal education is to prepare students for future learning when there is no longer classroom supervision. To continue learning, people need to learn to self-assess their own progress and understanding. We are investigating the social basis of self-assessment for learning. Specifically, we hypothesize that, under identifiable conditions, assessing others can support self-assessments that promote content learning plus the ability to develop self-assessment strategies that can be applied in the future. In this project, we take advantage of Teachable Agent technologies where students learn by teaching computer agents through the use of well-formed visual representations. Teachable Agents, using simple artificial intelligence techniques, can then reason based on what they have been taught. This creates optimal conditions for self-assessment, because students' assessments of their agents' performance is also an assessment of their own knowledge. The work occurs in the context of teaching the key ecosystem concepts of interdependence and balance to middle school students. Students will first create Teachable Agents that are linked to their curriculum on pond and river ecosystems, and use this learning experience to create a new Teachable Agent that can sustain multiple fish in a home aquarium system. Students will also use the Teachable Agents in a new homework model that leverages current trends in home computer use and connects learning in formal and informal settings; students log on, chat with one another, and their agents interact with another in an on-line virtual environment. Overall, the proposed project joins three important strands of research assisted by advanced technology tools: The learning of dynamic processes in science; the social basis of self-assessments for learning; and, the improvement of connections between formal and informal learning settings.

This award enables graduate student participation and the inclusion of invited speakers in the 18th International Workshop on Principles of Diagnosis (DX'07), Nashville, Tennessee. The workshop is an annual event focused on theories, principles, and computational techniques for diagnosis, monitoring, testing, reconfiguration, fault-adaptive control, and repair of complex systems, as well as applications to industry-related problems. NSF funding amplifies the educational impact of the meeting by enabling graduate students to participate actively in a scientific and technical forum that addresses an important facet of high-confidence, survivable systems.

Recent advances in artificial intelligence and robotics are confronting individuals of all ages with a series of category-defying entities that combine features of living and nonliving things. As such, these entities increasingly challenge people's basic understanding of mind and intelligence. The goal in this project is to explore adults' and children's beliefs about a range of living and mechanical agents, and to test how these beliefs affect people's ability to track, remember, and understand mechanical agents in two specific computer interfaces. First, it will explore a computer interface designed to allow a human operator to interact with and control a set of semi-autonomous robots. The second environment will be a teachable agent system in which middle school children learn about complex science phenomena, such as river ecosystems, by actively teaching an animated software agent.

This project represents one of the few research programs to empirically test people's understanding of living and artificial agents, and it will employ a conceptual framework that starts with naïve understandings of mind (e.g. "Theory of Mind") and applies them to engineered environments where these understandings are used. This framework describes the conditions under which participants apply different agent concepts, and can help understand how these beliefs might change over time as people interact with novel agents. Although the framework is not yet a complete theory, it represents a broadened approach to reasoning about both typical and novel living and mechanical agents that goes beyond existing dual-process models of Theory of Mind. These experiments also make links between concepts about agents, and the deployment of these concepts in realistic high-load perceptual tasks, so they can make an important contribution to our basic understanding of how knowledge affects vision.

The findings from this project may have important implications for educating both children and adults to deal with novel intelligent decision making technologies that move beyond the simple command-and-response cycle inherent to most current computer applications. Previous research by the PIs has already documented ways in which different people vary in their approach to these technologies (e.g. older and younger adults seem to have subtly different beliefs about the nature of computer intelligence), so this project may help improve the accessibility of novel agent-technologies to a wide range of different populations. More generally, because this research uses interactive educational tools and realistic robot-command systems to explore agent-understanding, it has the potential to improve user interfaces supporting social learning environments that focus on self-regulated learning, and that facilitate the effectiveness of human-machine emergency response teams. These technologies confront users with challenges to their most basic understandings of intelligence and thinking, and our research has the potential to guide both children and adults as they become successful users and creators of the interactive technologies of the future.

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).

Over the past decade, new and exciting technologies have created opportunities for developing rich open-ended learning environments that combine a number of different learning paradigms and resources. Students can complete quests in game environments, engage in inquiry, interact with virtual agents, run science simulations, take quizzes, access the web, and more generally make choices about different learning activities. These choices can be extremely informative about student learning and diagnostic of what students can learn once they leave highly scripted curricula. By employing machine learning methods, such as hidden Markov Models and other sequence analysis algorithms to analyze student choices in relatively open learning environments and determine whether students are showing (sub) optimal behavior patterns, the environment can then adapt intelligently by encouraging (alternative) choices and better learning behaviors. The primary hypothesis is that helping students develop the metacognitive abilities to make learning choices will have strong effects on their subsequent abilities to learn in the future in unstructured and unsupervised but resource-rich environments. The goals for this project are to create: a) Choice adaptive intelligent learning environments and computational methodologies that help students develop strategies to enable them to learn on their own; b) Novel automated assessment tools for both teachers and students that link choice and learning behaviors to learning performance; and c) Research studies that will establish whether our interventions that combine choice with guidance is beneficial for both strong and weak learners in science domains.

The broader impacts of this work span multiple dimensions. First, it provides an encompassing computer science framework for bringing together a number of technology-rich, interactive environments that are proliferating for education into a common choice filled and adaptive architecture. Second, this choice-based framework provides a paradigm shift in that tracking and theorizing about choice is applied in the context of learning, which, in the past, has been dominated by characterizations of the knowledge construct. Characterizing learning by choice not only connects learning research to a larger body of social science research, it is also a fundamentally new way to characterize and guide learning that is closer to the goal of much instruction, namely intelligent future choice. Third, the computer environment should permit the collection and analysis of large log files by many researchers, and conceivably lead to a new database of common choice patterns and their effects on learning. We will create the framework that enables others to incorporate intelligence into their virtual worlds and help achieve these proposed outcomes.

This proposal seeks funding to support the travel of advanced doctoral students so that they can participate in the 15th International Conference in Artificial Intelligence in Education and in the Young Researchers Track (YRT) that is part of that conference. The Young Researchers Track provides a forum for Ph.D. students to present and discuss their work with mentors from outside of their home institutions and to meet peers with similar interests. Mentoring happens in two ways -- as part of the Young Researchers Track, where they will be encouraged and offered feedback by three mentors, and through an individual mentoring program where each doctoral student will be matched with a mentor with expertise in the student's research area. Mentors will be senior AI and ED researchers.

Participants in the Young Researchers Track are members of the next generation of cyberlearning researchers. Participation in the proposed program will supplement their education at their home institutions and help prepare them to be leaders in transforming education through the use of learning technologies.

This development and research project from Vanderbilt University, Facet Innovations, and Filament Games, designs, develops, and tests a digital game-based learning environment for supporting, assessing and analyzing middle school students' conceptual knowledge in learning physics, specifically Newtonian mechanics. This research integrates work from prior findings and refines computer assisted testing and Hidden Markov Modeling to develop a new methodology to engage students in deep learning while diagnosing and scaffolding the learning of Newtonian mechanics.

The project uses a randomized experimental 2 x 1 design comparing a single control condition to a single experimental condition with multiple iterations to test the impact of the game on the learning of Newtonian physics. Using designed based research with teachers and students, the researchers are iteratively developing and testing the interactions and knowledge acquisition of students through interviews, pre and post tests and stealth assessment. Student learner action logs are recorded during game-play along with randomized student interviews. Students' explanations and game-play data are collected and analyzed for changes in domain understanding using pre-post tests assessment.

The project will afford the validation of EGAME as an enabler of new knowledge in the fields of cognition, conceptual change, computer adaptive testing and Hidden Markov Modeling as 90 to 300 middle school students learn Newtonian mechanics, and other science content in game-based learning and design. The design of this digital game platform encompasses a very flexible environment that will be accessible to a diverse group of audiences, and have a transformational affect that will advance theory, design and practice in game-based learning environments.

This project's investigation is at the nexus between scientific thinking, computational thinking, modeling as an investigative endeavor, and visual programming tools. The PIs are infusing middle-school science with efforts to promote computational thinking, doing that through making modeling a more significant part of science activities. The modeling experiences learners have become progressively more complex throughout each module and more complex across modules, with the increases in complexity informed by complexities of becoming a computational thinker. Modeling and computational thinking are foregrounded in each module, with each becoming more fluid over time as a result of the repetition of increasingly complex modeling experiences in a variety of situations, all of which build on each other. The mental model building, computational thinking, modeling, and science education literatures all inform the endeavor. The technological innovation includes creating and refining a modeling environment appropriate to middle schoolers, including an appropriate visual programming language. Research questions address issues in learning computational thinking in the context of learning to model and use models for investigation (and vice versa) and trajectories towards competency in computational thinking and modeling as their research questions.

Computational thinking is becoming a more and more important required expertise of scientists -- both those who work at the high levels of computational science and engineering and those who support them and apply computational science. In addition, as computation becomes more and more ubiquitous in a whole variety of disciplines and workplace responsibilities, the rest of the population, too, needs to be more expert at computational thinking and at using computational tools. Infusing computational thinking and the use of computational tools into the curriculum in appropriate ways is the right way to promote this cross-cutting expertise. Science is one place in the curriculum where computational thinking can easily be integrated, and doing so not only holds the promise for readying more of the population for careers and jobs that require computational thinking and use of computational tools but also making middle school science more exciting to more of the population.

This project uses broadband communication to support a diverse community of STEM learners, including K-12 students, their teachers and parents, school administrators, community organizers, city planners, and university faculty and graduate students. The project will build a community-situated, challenge-based, collaborative cyber-learning environment (C3STEM) that harnesses computational thinking, modeling, simulation, and challenge-based learning to support STEM learning in the context of a city traffic planning. C3STEM will use the traffic domain as the source for real-world problems for STEM education by developing projects through which high school students (from at least two schools) will collaboratively address problems of traffic congestion and safety in urban and suburban environments. Classroom-based student groups will work with their classroom teacher, traffic experts, and the Vanderbilt researchers to analyze real traffic data from regions near their homes and schools. The collaborative effort will lead to the students gaining an understanding of the traffic patterns, which will support the students developing agent-based models that align with the observed patterns in their section (e.g., traffic congestion along selected thoroughfares at different times of day and the effects of stoplights and interstate on/off-ramps). This process will teach fundamental concepts of data analysis and computer-based modeling and will motivate other curriculum-related mathematics and science lessons such as, Physics concepts, such as inertia, speed and acceleration, and Mathematical concepts, such as algebra, calculus, probability, and statistics. The project will leverage GENI and broadband infrastructure to provide real-time collaborative sessions and large-scale data transfers.

Broader Impact:
Via the combination of broadband communication, live data from community resources, computational modeling and collaborative learning, the products of student labor will not simply be for a grade, but their work will be regarded as an contribution to the community. This represents a cultural shift, which not only puts STEM in the foreground of active citizenship, but also raises the stature of teachers and K-12 education. The project results will be not only disseminated through journals and conferences but student projects will be displayed on a Web-based 'science-fair' and the PIs will create a Wikibook that archives the lesson plans and their evaluation. This text can be added to and revised by the community, as well as informing and supporting future community-situated, challenge-based programs.

The Cyberlearning and Future Learning Technologies Program funds efforts that support envisioning the future of learning technologies and advance what we know about how people learn in technology-rich environments. Development and Implementation (DIP) Projects build on proof-of-concept work that shows the possibilities of the proposed new type of learning technology, and PI teams build and refine a minimally-viable example of their proposed innovation that allows them to understand how such technology should be designed and used in the future and that allows them to answer questions about how people learn, how to foster or assess learning, and/or how to design for learning. An important issue in education is helping learners understand scientific phenomena, especially those that are too small or large, fast or slow, dangerous or inconvenient to experience and manipulate first hand. A way of helping learners experience such phenomena is through modeling -- building a model of the phenomenon or process and then manipulating it to see what happens in different circumstances. Computer tools are available for creating such models, but model building, though very useful for learning, is a complex and difficult task for many learners. In this team's Cyberlearning Exploration (EXP) Project, they developed what looks like a promising way to help middle school students learn to build computational models of scientific phenomena. They designed a visual language for expressing models and showed how progressing from less to more sophisticated models across several phenomena could help middle schoolers not only learn targeted science content but also learn how to design and build models and how to interpret and learn from models. In this follow-on project, they build on that approach, aiming to extend the technology to cover more sciences, automate some of the help teachers provide to students as they engage in model building and interpretation, systematically study the challenges learners face in learning through model building and interpretation, and identify pedagogical approaches that will foster successful learning in these circumstances.


The goal of this proposal is to improve middle schoolers' computational thinking and scientific modeling capabilities in parallel with each other and in a way that prepares young learners for the computational sciences of the future. The PIs' earlier Cyberlearning EXP project explored the potential of using what is known about learning to be a computational thinker to guide sequencing of activities for fostering model-building and interpretation capabilities in science. The sequencing they proposed has middle schoolers building models of phenomena that they gradually make more sophisticated over the course of a science unit, then in the next unit, repeating that sequencing again, but with different content requiring more sophisticated or different modeling practices. They designed a language for model specification that they hoped would foster computational thinking and model-building capabilities. CTSIM, the software that supports the approach, is a visual programming platform for modeling scientific modeling that includes discipline-specific constructs, provides a scalable architecture that seamlessly weaves together model construction, simulation, testing, experimentation, and verification, is usable across science domains, and connects to NetLogo, which runs the simulations. In this project, the team will extend the technology to cover more sciences, add adaptive scaffolding based on what they saw teachers providing to help students build and learn from models, identify pedagogical approaches for promoting learning from model building, and systematically study the challenges students face and how to address those challenges. Their research questions focus on students' abilities to simultaneously learn science content and computational thinking skills involved in modeling, the ins and outs of using linked representations, the kinds of feedback needed for scaffolding the approach, and scaling issues for teachers.

Innovative ideas are needed to improve the relevance of STEM education and student engagement. It is known that student motivation and engagement can be improved by challenging them with open-ended, real-world problems and having them work in multi-disciplinary teams. The proposed research will achieve these objectives and impart STEM knowledge and 21st century skills by connecting course content and assessments to deep STEM learning and real-world engineering problem solving. Using cloud-based software infrastuctures, students can use the proposed C3STEM environment to collaborate with their peers on knowledge intensive tasks, conduct joint experiments, and solve complex problems by decomposing them into smaller, more manageable tasks. This approach will improve the technological competence of students and help them to develop into global leaders.

The project will research, design and validate new technologies that will enable ubiquitous and collaborative STEM education. The team will design new software systems by exploiting the Cloudlet and Locavore patterns in conjunction with Cloud Computing. New techniques for Software Defined Networking (SDN) will be designed and validated to dynamically create and manage network bandwidth in support of the envisioned applications. Resource allocation, scalability and Quality of Service issues will be addressed in the integrated context of SDN blended with Cloudlets. The investigation will address insights and scientific foundations that inform new directions in Cloud Computing and Network Virtualization as applied to collaborative engineering problem solving. The team will design STEM curricular units using real-world applications, such as traffic flow in city streets and discrete manufacturing systems that are directed at high school and undergraduate students. The investigators will design, instrument, and validate next-generation test beds for remote access using multiple mobile devices and computing platforms for performing experiments and problem-solving by remote teams. For example, high school student teams in Nashville, TN will be able to collaborate with graduate and undergraduate students at Vanderbilt University and interact with students in Akron, OH while using a manufacturing test bed located at the University of Akron. The distributed team will be able to collaboratively collect real-time data from the test beds via mobile devices, program new behaviors, validate their designs and engage in solving complex problems of real-world significance.

BIGDATA: Infrastructure and Analytics for Data Intensive Research in Open-Ended Learning Environments

Data science techniques have revolutionized many academic fields and led to terrific gains in the commercial sector. They have to date been underutilized in solving critical problems in the US educational system, particularly in understanding Science, Technology, Engineering and Mathematics (STEM) learning and learning environments, broadening participation in STEM, and increasing retention for students traditionally underserved in STEM. The goals of the Directorate for Education and Human Resources (EHR), through the EHR Core Research program, for the Critical Techniques and Technologies for Advancing Foundations and Applications of Big Data Science & Engineering (BIGDATA) program are to advance fundamental research aimed at understanding and solving these critical problems, and to catalyze the use of data science in Education Research. As more open-ended learning environments are being employed in schools, it is critical to understand how they affect learning, motivation and engagement in STEM. The most widely used methods in educational research are inadequate for addressing these questions because they do not address the scale and complexity of the data provided from these environments. This Early Concept Grant for Exploratory Research (EAGER) will advance the understanding of how promising new technology environments affect these outcomes by developing a data repository and open-source analytical tools. The PI will show proof of concept by using these tools with data from three different types of online learning environments.

The main goal of the proposal is to develop an open-ended learning environment (OELE) educational dataset repository and novel integration and analysis techniques that tap the potential of both theory driven and bottom-up data mining for understanding learning in OELEs. In addition, the Principal Investigator will develop an analysis environment incorporating common tools to support researchers and practitioners in conducting analyses. The proposed data integration challenge is ambitious and risky. However, the PI has the expertise to complete the project. Both the exploration of data science methods in open-ended learning environments and of the use of bottom up approaches to data analysis are greatly needed in educational research. Most computer based learning environments are more open ended than those for which the majority of the methodologies used in educational data mining and learning analytics have been developed. The use of bottom up methods of data mining has produced phenomenal results in the commercial sector and in academic fields such as biology. There are almost no other similar efforts currently underway in education in the academic setting, and most in the commercial sector are not open source projects.

This award is supported by the EHR Core Research (ECR) program. The ECR program emphasizes fundamental STEM education research that generates foundational knowledge in the field. Investments are made in critical areas that are essential, broad and enduring: STEM learning and STEM learning environments, broadening participation in STEM, and STEM workforce development.

The Cyberlearning and Future Learning Technologies Program funds efforts that support envisioning the future of learning technologies and advance what we know about how people learn in technology-rich environments. Cyberlearning Exploration (EXP) Projects explore the viability of new kinds of learning technologies by designing and building new kinds of learning technologies and studying their possibilities for fostering learning and challenges to using them effectively. This project will lay the groundwork necessary for incorporating eye movements into cyberlearning. Although hardware and software solutions are rapidly advancing the ability to detect and track cyberlearners' eye movements, the scientific understanding of the link between these eye movements and actual learning remains tentative. This issue is particularly important because research demonstrates surprising limits to the visual information that people take in: Even when it can be demonstrated that they have looked at something, this is no guarantee that learners gain knowledge of what they have seen. This project will address this problem in two ways. First, the researchers will develop a cognitive theory that can help specify how eye movements reveal what cyberlearners have absorbed when they view and interact with technology-based learning systems. Second, the researchers will develop a novel software application that helps cyberlearning content creators to incorporate assessment of eye movements into their practice. These projects will converge not only to develop cognitive theory that can help cyberlearners achieve more effective interactions, but also to enrich cognitive theory with input from real-world cyberlearning practitioners who struggle every day with the need to understand the sometimes confounding link between showing a learner something and learners' actual ability to understand and remember what they have seen.

In particular, the investigators hypothesize that the link between fixation patterns and learning is mediated by visual modes that vary the relationship between concrete coding of visual properties and abstract focus on causal relationships and the goals of actions. The project will include experiments in which learners have their eyes tracked while they view a screen-captured information technology lesson. Some learners will be induced to deploy an "encoding" mode in which they focus on the specific sequence of steps needed to complete the task, while other learners will view the same materials using a "causal" mode in which they focus on the concepts underlying the lesson. Initial research has demonstrated significant differences in fixation patterns in these tasks (the strongest of these is that learners follow the instructor's mouse movements more closely in the encoding mode), and the current project will test whether these modes are associated with different patterns of visual and conceptual learning. The project will leverage these results by incorporating mode-revealing analytics into a novel software application that allows content creators to record screen-capture videos of their lessons while recording their own eye movements. In addition, a panel of viewers will be equipped with their own eye trackers and will view the content creators' lessons. Viewer eye movements will be returned to content creators who will be able view fixation patterns in the application, along with analytics based on findings from the visual mode experiments. The prototype system will be integrated with an existing learning technology, courseware for computer science education titled "Betty's Brain," and deployed in both formal and informal learning environments, including the Nashville Adventure Science Center.

The objective of this research is to understand and improve the resource coordination and dispatch mechanisms used by first responders in smart and connected communities. In prior art, as well as practice, incident forecasting and response are typically siloed by category and department, reducing effectiveness of prediction and precluding efficient coordination of resources. This research project provides a unique opportunity to study the problem by integrating both the data and emergency resources from distinct urban agencies in the City of Nashville along with other widely available data such as pedestrian traffic, road characteristics, traffic congestion, and weather. This will allow development of models for anticipating heterogeneous incidents, such as distinct categories of crime, as well as vehicular accidents. With these models we can develop decision support tools to optimize both resource allocation and response times. These tools will help the emergency responders determine which units to dispatch (police, fire, or both) in order to minimize expected response time, and what equipment is most appropriate, taking into account the time, location, and nature of incidents, as well as those predicted to occur in the future. Ultimately, the methods developed in this research can be applied to other domains where multi-resource spatio-temporal scheduling is a challenge.

The technical aspects of this project will require us to develop methods for solving the algorithmic challenge related to continuous-time forecasting of spatio-temporal time series of heterogeneous incidents. In tackling the forecasting task, we will develop methods to cluster incidents taking into account multiple features, and use the resulting groupings to develop distinct continuous-time models that forecast incident occurrence distributions based on survival analysis. The optimization framework, in turn, requires a scalable solution for integrated spatio-temporal allocation of heterogeneous emergency responders, making use of developed integrated forecasting methods. The proposed optimization methods will transform the incident response problem into a transportation problem with heterogeneous resources, which can be formalized as a network-flow linear program, augmented to account for heterogeneity in the resources and incidents that these resources can address. The developed solutions will be made available to the community for maximal dissemination. This research has the potential to impact actual operational planning at the Metro Nashville Police Department and Nashville Fire Department, by optimally coordinating responses. Broader impacts also include involvement in educational activities, including STEM-related projects for High School students at the School for Science and Math at Vanderbilt, undergraduate and graduate teaching, and active engagement of undergraduates and graduates in research.

This project aims to enhance theory and measurement of students' self-regulated learning (SRL) processes during science learning. SRL refers to learning that is guided by metacognition (thinking about one's thinking), strategic action (planning, monitoring, and evaluating personal progress against a standard), and motivation to learn. The project will accomplish this by developing a technology-based framework which leverages human expert judgment and machine learning methods to identify key moments during SRL and analyze these moments in depth. The project uses an existing science learning environment, Betty's Brain, that combines learning-by-modeling with critical thinking and problem solving skills to teach complex science topics. The environment is designed to have the student teach Betty science topics using concept maps (the critical elements of the science) and reading materials provided by the environment. A critical component of this project is to determine when a student using the system needs help. Using SRL as a basis, the additions to Betty's brain will identify key points in the SRL processes of metacognition, strategic action, and motivation. Some of these points can be determined automatically in recognizing key points while others require human intervention to recognize key points and then to determine what actions should be taken to enable the student. This information will be recorded and the experiences then are used to update the automatic identification of key points.

The project's main intellectual merit is in integrating the power of data mining to rapidly sift through large amounts of data to find key inflection or change points in student reasoning and strategies, with the power of human beings to deeply understand other humans' SRL processes. This measurement framework, with the accompanying detectors of inflection points in students' SRL in online learning, has the potential to transform science learning and teaching in K-12 settings by providing insights into how SRL unfolds during learning through the interactions between affect, engagement, cognition and metacognition. Those insights will be used to extend an existing theory of SRL, increasing its richness, specificity, and predictive power. Self-regulated learning is important to student success, both in K-12 education and during life-long learning afterwards. Better understanding of SRL processes will support the development of computer-based science learning environments, such as Betty's Brain, with the capability to better support students' learning of SRL skills and strategies in science classrooms. By studying these issues within the diverse population of urban students who currently use Betty's Brain, success of the project will increase the relevance of SRL to the full diversity of America's learners. The project's software will be available through the portal www.teachableagents.org.

Computing and computational thinking (CT) are an integral part of everyday practice within modern fields of science, technology, engineering, and math (STEM). As a result, the STEM+Computing Partnerships (STEM+C) program seeks to advance new multidisciplinary approaches to, and evidence-based understanding of, the integration of computing in STEM teaching and learning, and discipline-specific efforts in computing designed to build an evidence base for teaching and learning of computer science in K-12, including within diverse populations. Integrating CT into core science instruction addresses practical constraints in K-12 education, in that there is no room in the curriculum to teach it directly to everyone. But, more importantly, integrating CT into core science provides a synergistic opportunity to deepen instruction in both. This project investigates the synergistic learning of physics and CT concepts and practices through students' construction of and interaction with computational models that visually represent physical systems. Led by investigators at Vanderbilt University, the project team includes computer scientists, physicists, education developers, and learning scientists SRI International, Stanford, and Salem State University. The project will develop, implement, and study an innovative programming environment, a multi-week computational physics curriculum, and new assessments that are focused on physics concepts of force and motion and CT practices involved in computational modeling. Guided by the programming environment and the curriculum, learners construct models that represent physical systems, analyze and explain model behaviors, and then use models for solving problems. These processes support their abilities to think and act like a scientist as they explore and learn about both the computational and physical systems and phenomena. Assessments will be developed to measure CT-infused physics learning that is targeted in the curriculum, but also what CT learners apply to new physics topics and problem-solving situations they encounter. The educational program will address specific needs of high school students and teachers with regard to relevant disciplinary content, practices, and computation as specified in Next Generation Science Standards, the AP Computer Science Principles, and recent consensus frameworks for computational thinking in STEM. Approximately 450 students will be involved with and benefit from the project in four diverse high school settings. The diverse nature of the participating schools will both engage a demographically diverse student population in STEM and help the project achieve significant broader impacts, by assuring that the findings and products developed reflect the needs of a broad diversity of people and places. The project will develop new educational technologies, curriculum materials, and assessments for integrating physics and computation that will be broadly usable in high school physics and computer science courses.

This project will investigate a method of broadening access to CT through the integration of computational modeling and problem solving in secondary physics courses. Through constructing computational models that represent complex physical systems, students will learn key concepts of Newtonian physics and CT practices of problem representation, abstraction, decomposition, composition, and verification. The project will produce a new programming environment that is optimized for modeling physics systems and phenomena, that facilitates collaborative modeling and problem solving, and that diagnoses and responds to users' learning activity with adaptive scaffolds. Three standards-aligned, problem-oriented computational physics units will be developed and used in conjunction with the programming environment. Evidence-centered design will be used to develop and validate assessments that measure CT-infused physics learning that is targeted in the units. A unique set of assessments will be developed independent of the modeling environment to measure whether and what CT students spontaneously transfer to new physics problems and learning situations. The curriculum and assessments with be co-developed by researchers and four teachers from diverse high school settings in Tennessee and California. Quantitative and qualitative analyses of student assessments, surveys, work products, computer-use logs, and videotaped tasks will be used to determine the effectiveness and broad utility of the approach - and its component parts - for integrating physics and computing. By tackling the challenge to align offline and online measures of students' learning and behaviors, the project will generate deeper understanding of how students learn, the difficulties they face, and the promise of adaptive scaffolds for improving learning. The research will also elucidate the potential of explicit CT frameworks for preparing students' for future physics learning and problem solving. The project will provide the field with a strong foundation for designing learning technologies that integrate science and computational modeling. The research findings will be shared with the project team members' respective communities in computer science, technology education, cyberlearning, physics education, science education, and teacher education through papers in peer-reviewed journals and conference presentations. Efforts will be made to disseminate to teachers through practitioners workshops, conferences, and journals. A post-doctoral fellow and three graduate students will be trained through this project.

Intelligent, interactive, and highly networked machines -- with which people increasingly share their autonomy and agency -- are a growing part of the landscape, particularly in regard to work. As automation today moves from the factory floor to knowledge and service occupations, insight and action are needed to reap the benefits in increased productivity and increased job opportunities, and to mitigate social costs. Such innovations also have significant implications and potential value for lifelong learning, skills assessments, and job training/retraining in an environment in which workforce demands are changing rapidly. The workshop supported by this award will promote the convergence of cognitive psychology, learning sciences, data science, computer science, and engineering disciplines to define key challenges and research imperatives of the nexus of humans, technology, and work with focus on human affect, motivation, metacognition, and cognition during learning and problem solving. Convergence is the deep integration of knowledge, theories, methods, and data from multiple fields to form new and expanded frameworks for addressing scientific and societal challenges and opportunities. This convergence workshop addresses the future of work at the human-technology frontier.

The specific focus of this multi-phased workshop approach is to advance fundamental understanding of how to collect and analyze multimodal, multichannel sensor on human affect, motivation, metacognition, and cognition during learning and problem solving, and effectively integrate this data into actionable educational interventions in advanced learning technology environments (e.g., intelligent tutoring systems). The impacts of this research extend to a diverse range of learning environments, and job training and retraining opportunities. A multi-phased workshop approach will be used to explore the implications in multiple job sectors, and the outcomes will be broadly disseminated across geographic and disciplinary boundaries.

The Future of Work at the Human-Technology Frontier (FW-HTF) is one of 10 new Big Ideas for Future Investment announced by the National Science Foundation. The FW-HTF cross-directorate program aims to respond to the challenges and opportunities of the changing landscape of jobs and work by supporting convergent research. This award fulfills part of that aim. Effective decision making by power grid operators in extreme events (e.g., Hurricane Maria in Puerto Rico, the Ukraine cyber attack) depends on two factors: operator knowledge acquired through training and experience, and appropriate decision support tools. Decision making in electric grid operation during extreme adverse events directly impacts the life of citizens. This project will augment the cognitive performance of human operators with new, human-focused decision support tools and better, data-driven training for managing the grid especially under highly disruptive conditions. The development of new generation of tools for online knowledge fusion, event detection, cyber-physical-human analysis in operational environment can be applied during extreme events and provide energy to critical facilities like hospitals, city halls and essential infrastructure to keep citizens safe and avoid economic loss for the Nation. Higher performance of operators will improve worker quality of life and will enhance the economic and social well-being of the country. The project's training objectives will leverage existing educational efforts and outreach activities and we will publicize the multidisciplinary outcomes through multiple venues.


The proposed project will integrate principles from cognitive neuroscience, artificial intelligence, machine learning, data science, cybersecurity, and power engineering to augment power grid operators for better performance. Two key parameters influencing human performance from the dynamic attentional control (DAC) framework are working memory (WM) capacity, the ability to maintain information in the focus of attention, and cognitive flexibility (CF), the ability to use feedback to redirect decision making given fast changing system scenarios. The project will achieve its goals through analyzing WM and CF and performance of power grid operators during extreme events; augmenting cognitive performance through advanced machine learning based decision support tools and adaptive human-machine system; and developing theory-driven training simulators for advancing cognitive performance of human operators for enhanced grid resilience. A new set of algorithms have been proposed for data-driven event detection, anomaly flag processing, root cause analysis and decision support using Tree Augmented naive Bayesian Net (TAN) structure, Minimum Weighted Spanning Tree (MWST) using the Mutual Information (MI) metric, and unsupervised learning improved for online learning and decision making. Additionally, visualization tools have been proposed using cognitive factor analysis and human error analysis. We propose a training process driven by cognitive and physiometric analysis and inspired by our experience in operators training in multiple domain: the power grid, aircraft and spacecraft flight simulators. A systematic approach for human operator decision making is proposed using quantifiable human and engineering analysis indices for power grid resiliency.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Ubiquitous access to mobile and web technologies enables the public to share valuable information about their surroundings anywhere and anytime. For example, during an emergency or crisis people report needs from affected areas via social media as an alternative to the traditional 911 calls. This can be valuable information for a range of emergency service officials. However, the utilization of this data poses several computational challenges as it is generated in real time, is heterogeneous, highly unstructured, redundant, and sometimes unreliable. This project innovates in two specific directions to alleviate the challenges associated with large, streaming datasets during emergencies: (1) The project investigates new summarization approaches to handle noisy, unstructured data streams from multiple web sources in real time while accounting for the possibility of untrustworthy information, so that they can be fed into decision support systems of public services in a structured and machine-readable format. (2) The project develops and validates robust decision support systems for allocating critical resources to needed areas based on the structured summary reports. The evaluation plan includes collaboration with emergency responders and the communities they serve. The broader impacts of this research include the design of a generic methodology to extract, integrate, and summarize structured information from big data streams on the web for helping public services of future smart cities. The research team plans to share simulated datasets with an open source system for real-time decision support during emergency response exercises. This can assist in workforce training and also, help design novel educational projects of data science for social good.

Formally, this research project investigates the theories behind a novel knowledge representation called Uncertain Concept Graph. The graph contains heterogeneous nodes based on key concepts of an application domain (e.g., regions, incidents, and information sources during a disaster). The graph has heterogeneous edges connecting these concept nodes, based on the inference of concept relationships using the extracted information from data streams (e.g., Twitter and news sources). The structure of the graph evolves over time and both nodes and edges can be added, deleted, or updated. An equivalent Bayesian Network is derived from the Uncertain Concept Graph describing the dependencies between the events captured in the graph at a given time instance. Based on the relationship edges in a graph state and the constructed Bayesian Network, an action recommendation system is created to support an application domain task (e.g., dispatching ambulance resources to incident-specific regions). To ensure robustness, this project develops and validates a novel anomaly identification and diagnosis approach using mode similarity to assess the correctness of current state of concept nodes and their relationships in the Uncertain Concept Graph at any time. The research team uses historical datasets of recent disasters to construct the graph and develop a demo system for domain evaluation, in order to recommend actions in emergency response for the city emergency services. The investigators are including the lessons learned and methodologies developed in their respective course curriculums.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Each year in the United States, approximately 70,000 new adults on the autism spectrum will seek employment. At the same time, employers in technology, finance, healthcare, and many other critical job sectors seek highly skilled and highly trained individuals to fill specialized positions. With support from the DRK-12 Program in the Division of Research on Learning and the NSF 2026 Fund Program in the Office of Integrated Activities, this research will investigate new tools and methods for matching individual job-seekers on the autism spectrum to employment opportunities that leverage their unique cognitive skills, with a focus on visuospatial cognitive skills. Numerous jobs require strong visuospatial cognitive skills, such as visual inspection and quality control, process monitoring, document review, surveillance, software testing, and data visualization, to name a few. Many people on the autism spectrum show strengths in visuospatial cognitive skills, but these strengths are not fully understood, including how they differ from person to person and how they map onto workplace-relevant capabilities. Understanding visuospatial cognitive skills in individuals on the autism spectrum or other neurodiverse conditions has high potential impact for enhancing the neurodiversity of the workforce by enabling more effective programs for the recruitment, selection, and retention of such candidates in the public and private sectors.

This NSF2026 EAGER project enriches the NSF2026 Idea Machine winning entry Harnessing the Human Diversity of Mind. It seeks to develop and evaluate integrated, AI-enabled technologies for measuring a person?s visuospatial cognitive skills in new ways and then using these measurements to predict performance on workplace-relevant tasks. The research conducted during this two-year project will include conducting a large pilot study with individuals on the autism spectrum and neurotypical individuals, in which participants will be given several visuospatial tests, and detailed data about their actions will be recorded using sensors such as eye trackers and cameras. Then, data mining and machine learning techniques will be used to extract meaningful patterns from these rich streams of behavioral data, and analyses will be conducted to examine how these patterns in foundational behaviors map onto individual skills and interests in realistic, workplace-relevant activities. This research will also gather and analyze detailed feedback from industry partners to identify specific job types and sectors that would benefit from recruiting employees who are strong in visuospatial cognitive skills. In addition, this project will involve neurodiverse students and staff in many of its activities, in particular by involving graduate trainees supported by the NSF Research Traineeship in Neurodiversity Inspired Science & Engineering (NISE) and by leveraging the skills of neurodiverse interns at the Frist Center for Autism & Innovation at Vanderbilt University's School of Engineering.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Computational thinking (CT) is the foundation of modern competency in a multitude of STEM-related fields. Learning strategies influence the way a student processes information and learns, which in turn, requires the proper control and regulation of their cognitive processes. More specifically, an effective learning strategy requires students to have adequate descriptive, procedural, and conditional knowledge of the strategies they apply. This project will study the learning processes of middle school students when they are involved in learning science using a computational modeling approach. The research that will be performed will help us to better understand how to effectively integrate CT and computing into the K-12 science curricula. To examine how learning evolves, and the difficulties students face in developing and applying synergistic processes, a task-oriented framework will be developed to analyze student learning behaviors as they work on modeling and problem-solving tasks.

This project will use adaptive scaffolding to synergistically teach scientific concepts with computational modeling to middle school students. Student learning behaviors will be analyzed using self-regulated learning theory. This project aims to make the following research contributions: (1) develop a framework for analyzing student learning behaviors and strategies as they are involved in their model building, model debugging, and problem solving tasks; (2) use a combination of students activity logs and eye tracking data to understand student cognitive and metacognitive processes as they work in the C2STEM environment; and (3) study the effectiveness of the adaptive scaffolding and feedback generation framework by analyzing how this helps students overcome their difficulties, and progress in their learning and problem solving tasks. An intelligent peer agent in an artificial intelligence system will be developed to achieve these goals.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.