|
Multinomial Processing Tree Model: New projects and implementations
Funding agency: National Science Foundation
Funding period: 2006-2009
Amount: $240,000 (Xiangen Hu, PI)
Effectiveness of Pedagogical Agents in Regulating Students’ Understanding of Science
Funding agency: National Science Foundation
Funding period: 2006-2009
Amount: $904,581 (Roger Azevedo, PI)
The Learning Kit-Theory and Cognitive Tools to Enhance Learning Skills and Support Life-Long Learning
Funding agency: Social Sciences and Humanities Research Council of Canada
Funding period: 2003-2007
Amount: $2,239,884 (Roger Azevedo, CoPI)
The Role of Self-Regulated Learning in Students’ Understanding of Science with Hypermedia
Funding agency: National Science Foundation (NSF) Early Career Award and Grant
Funding period: 2002-2007
Amount: $615,663 (Roger Azevedo, PI)
An In-depth Analysis of Expert Human Tutors
Funding agency: Office of Naval Research
Funding period: 2005-2007
Amount: $445,674 (Natalie Person, PI)
Tracking Multimodal Communication in Humans and Agents
Funding agency: National Science Foundation
Funding period: 2004-2006
Amount: $699,949 (Max Louwerse, PI)
iSTART: Interactive Strategy Training for Active Reading and Thinking
Funding agency: Institute of Education Sciences
Funding period: 2004-2007
Amount: $1,770,514 (Danielle McNamara, PI)
Why2000: A Tutor that Teaches Mental Models Using Natural Language Dialogs
Funded by: Office of Naval Research
Funding period: 2000-2005
Amount: $1,168,700 for University of Memphis (Art Graesser, PI; Kurt VanLehn, PI at University of Pittsburgh), additional $90,175 for graduate student fellowship
Computer Science Resources for Memphis Area High Schools
Funding Agency: National Science Foundation
PI/Co-PIs: Linda Sherrell (PI), Lee McCauley, and Sajjan Shiva
Funding Period: 2004 - 2006
Amount: $1,600,000
Developing AutoTutor for Computer Literacy and Physics
Funding agency: National Science Foundation
Funding period: 2001-2004
Amount: $1,274,075 (Art Graesser, PI)
Understanding in Science
Funding agency: National Science Foundation, Understanding in Science
Funding period: 2002-2004
Amount: $296,902 for University of Memphis subcontract (Art Graesser, PI; Jennifer Wiley, PI at University of Illinois at Chicago)
Developing and Testing a Computer Tool that Critiques Survey Questions
Funding agency: National Science Foundation
Funding period: 2000-2003
Amount: $205,990 (Art Graesser, PI)
Monitoring Emotions While Students Learn with AutoTutor
Funding agency: National Science Foundation
Funding period: 2003-2008
Amount: $1,250,000 (Art Graesser, PI)
A Web Site on Institutional Review Boards and the Ethical Use of Human Subjects
Funding agencies: Institute for Defense Analysis & ONR NAWCTSD
Funding period: 2000-2001
Amount: $950,000 (Art Graesser, PI; $750,000 through Thoughtware Corporation and $200,000 through The University of Memphis)
Question- Driven Explanatory Reasoning about Devices that Malfunction
Funding agency: Office of Naval Research
PI/Co-PIs: Art Graesser (PI)
Funding period: 1998-2000
Amount: $135,312
QUEST Questionnaire Evaluation Tool
Funding agency: United States Bureau of Census
PI/Co-PIs: Art Graesser (PI)
Funding period: 1998-1999
Amount: $58,512
Simulating Tutors with natural Dialogue and Pedagogical Strategies
Funding agency: National Science Foundation
PI/Co-PIs: Art Graesser (PI)
Funding period: 1997-2000
Amount: $900,000
Predicting Information Needed Prior to Request with Cognitive Models of the UserÕs Task Knowledge
Funding agency: Office of Naval Research
PI/Co-PIs: Art Graesser (PI)
Funding period: 1995-1996
Amount: $92,879
Program Gear- Up
Funding agency: U. S. Department of Education
PI/Co-PIs: Don Franceschetti (PI), Regina Hairston (Co- PI), Leonard Jamerson (Co- PI)
Funding period: 5 yrs, Ongoing
Amount: $900,000
NIH Bridge to Baccalaureate
Funding agency: National Institute of Health, Office of Minority Health
PI/Co-PIs: Don Franceschetti (PI), Regina Hairston (Co- PI)
Funding period: 9 of 10 yrs. Ongoing
Amount: $1,700,000
Tennessee Louis Stokes Alliance for Minority Participation in Science, Engineering, Mathematics and Technology
Funding agency: National Science Foundation
PI/Co-PIs: Art Graesser (PI)
Funding period: 5 yrs
Amount: $400,000
Scholarship for Computer Science, Engineering and Mathematics
Funding agency: National Science Foundation
PI/Co-PIs: David Russomanno (PI), Regina Hairston ( Co- PI)
Funding period: 2 yrs
Amount: $187,000
Promoting active reading strategies to improve students' understanding of science
Funding agency: National Science Foundation IERI
Funding period: 2000-2005
Amount: $3,196,000 (Danielle McNamara, PI)
Coh-Metrix: Automated Cohesion and Coherence Scores to Predict Text Readability and Facilitate Comprehension
Funding agency: Institute of Education Sciences
Funding period: 2002 -2005
Amount: $1,425,200 (Danielle McNamara, PI)
Data-Link Aircraft Communications: An Examination of Interference, Coherence, and Workload
Funding agency: National Aeronautics and Space Administration AMES Research Center
Funding period: 2001-2004
Amount: $365,000 (Danielle McNamara, PI)
Percolation Models of Phase Transition during Perceptual Processing
Funding agency: National Science Foundation /Quantum & Biologically Inspired Computing
Funding period; 2002-2005
Amount: $652,194 (Robert Kozma, PI)
A Multiresolution Wavelet-Fractal Approach to Abnormal Tissue Identification in Brain MR Images
Funding agency: The Whitaker Foundation
PI/Co-PIs: Khan M. Iftekharuddin (PI)
Funding period: 2002-2004
Self Organizing Ontogenetic Development for Autonomous Adaptive Systems
Funding agency: NASA/Ames Revolutionary Computing Systems
Funding period:
Amount: (Robert Kozma, PI)
Immunity-Based Intrusion Detection Systems
Funded agency: DARPA
Funding period: 2000- 2003
Amount: (Dipankar Dasgupta, PI)
Preliminary Research on Immunity-Based Computational Techniques
Funding agency: NSF SGER
Funding period: 2001-2002
Amount: (Dipankar Dasgupta, PI)
Eye Tracking While Answering Questions in Electronic Multimedia Environments
Funding agency: Office of Naval Research
PI/Co-PIs: Art Graesser (PI)
Funding period: 2001-2002
Amount: $120,000
An In-depth Analysis of Expert Human Tutors
Funding agency: Office of Naval Research
Funding period: 2005-2007
Amount: $445,674 (Natalie Person, PI)
Summary: We are proposing a three-year research project in which we will thoroughly examine the pedagogical, motivational, and conversational dimensions of expert human tutoring. The major goals of the project will be to (1) document the theoretical models and strategies of tutoring/teaching that guide the practices of expert human tutors, (2) identify the particular tactics, actions, or dialogue moves are employed by expert tutors, (3) contrast the practices of expert tutors with the practices of human tutors who are typically the focus of tutoring studies, and (4) produce a comprehensive roadmap of expert human tutoring.
There have been numerous studies in which the processes of human tutoring have been reported. However, these studies have not provided a clear picture of the processes that differentiate expert human tutors from other kinds of tutors. The underlying assumption in the tutoring literature is that tutoring is the most effective method of instruction and that expert human tutors are better than less skilled tutors. Given that the practices and products (i.e., student learning outcomes) of expert human tutors serve as benchmarks for many ITS designers, it is somewhat surprising that our knowledge of expert tutors is rather limited. Our knowledge is limited for several reasons. One reason is that most tutoring studies include typical or less experienced tutors rather than expert tutors. There are only a handful of studies in which expert tutors are the focus of the analyses. Other reasons have to do with the expert tutoring studies themselves. For example, in several expert tutoring studies, it is unclear how many tutors were included in the analyses. In other studies only one or two expert tutors were observed making it difficult to generalize the results to all expert tutors. It is also the case the there is considerable variability in what constitutes an expert tutor. Graduate students who are paid to work in a university tutoring program may be considered experts tutors in some studies but typical, non-expert tutors in others.
The expert tutors in the proposed research project will be twelve math and science tutors that have been recommended by the most prestigious private schools in the greater Memphis area. All of the tutors have long-standing relationships with the academic support offices that recommend them to parents and students. They are considered experts because they have considerable experience in one-to-one tutoring (at least five years), they having outstanding reputations as private tutors and are highly recommended by prestigious college-preparatory institutions, and students who work with these tutors show marked improvement in the subject areas for which they received tutoring.
Top
Tracking Multimodal Communication in Humans and Agents
Funding agency: National Science Foundation
Funding period: 2004-2006
Amount: $699,949 (Max Louwerse, PI)
Summary: This project investigates multimodal communication in humans and agents, focusing on two linguistic modalities - prosody and dialog structure, which reflect major communicative events, and three non-linguistic modalities - eye gaze, facial expressions, and body posture. It aims to determine (1) which of the non-linguistic modalities align with events marked by prosody and dialogue structure, and with one another; (2) whether, and if so when, these modalities are observed by the interlocutor; and (3) whether the correct use of these channels actually aids the interlocutor's comprehension. Answers to these questions should provide a better understanding of the use of communicative resources in discourse and can subsequently aid the development of more effective animated conversational agents.
The outcomes of our observations will be modeled on controlled elicited dialog. To assure robust information on the interplay of modalities, we control the base conditions, genre, topic, and goals of unscripted dialogs. An ideal task for this is the Map Task, where dialog participants work together to reproduce on one player's map a route preprinted on the other's. The two maps, however, are slightly different, so that each player holds information important to the other. This scenario triggers a highly interactive, incremental and multimodal conversation. In the proposed project a basic corpus of Map Task dialogues will be collected while recording spoken language, posture, facial expressions, and eye gaze. Hand gestures, discouraged by the task, will be recorded where they occur. These findings will be used in the Behavior Expression Animation Toolkit (BEAT) in order to augment the current intelligent system AutoTutor.
AutoTutor has been developed for a broad range of tutoring environments that coach the student in following an expected set of descriptions or explanations. The coach-follower roles in the Map Task scenario make it possible to easily change the scenario for AutoTutor. In a series of usability experiments interactions of dialog participants with AutoTutor will be recorded. These experiments allow us to record not only the participant's impressions, but also his or her efficiency (the time to complete map, latency to find named objects, deviation of the instruction follower's drawn route from the instruction giver's model), and communicative behavior (discourse structure, gaze, facial expressions, etc.).
The research resulting from this project will benefit a large variety of fields, including cognitive science, computational linguistics, artificial intelligence, and computer science. In addition, the integration of the modalities into a working model will advance the development and use of intelligent conversational systems.
Top
iSTART: Interactive Strategy Training for Active Reading and Thinking
Funding agency: Institute of Education Sciences
Funding period: 2004-2007
Amount: $1,770,514 (Danielle McNamara, PI)
Summary: Students in the U.S. show significant deficits in understanding and learning from complex written material. Thus, it is important to find ways to improve text comprehension for our nationÕs students. Our objective is to further develop and evaluate an automated reading strategy intervention called iSTART. This interactive strategy training system helps students understand and learn from difficult text by teaching them to use active, effective reading strategies. While reading strategies are particularly important for understanding textbooks, teachers are often inadequately prepared to provide students with reading strategy training.
We propose further developing iSTART such that it can be easily integrated into high-school classrooms. The current version of iSTART provides training to help students more effectively self-explain difficult text while reading. iSTART is based on experimental evidence that self-explanation coupled with reading strategy training (Self-Explanation Reading Training) improves comprehension and course grades. iSTART delivers reading strategy training using an interactive and adaptive format. Pedagogical agents interact with each other and with the student to increase active processing and participation. iSTART begins with instruction for self-explanation and reading strategies such as monitoring comprehension, making bridging inferences, and using prior knowledge and logic to understand the text. After this instructional phase, the student identifies strategies used within examples of self-explanations. The student practices by self-explaining science texts. The system evaluates the self explanations and provides feedback. Thus, iSTART includes four activities: instruction, modeling, strategy identification, and practice.
We propose augmenting iSTART such that it can be easily integrated into classrooms. This goal will be accomplished by: 1) increasing the number of texts and the range of domains available for reading strategy practice, 2) increasing the range of difficulty of the practice texts, 3) increasing the range of strategy interventions, and 4) creating an interface to facilitate teachersÕ use of iSTART. These modifications will enable the system to better adapt to student and teacher needs. We have proposed a systematic development and assessment plan which includes nine experimental studies. The final study will examine the augmented iSTART system coupled with the teacher interface in 30 classrooms. This project will guarantee a reading intervention that can then be experimentally tested in classrooms and finally scaled up to our NationÕs classrooms. Our ultimate goal is for iSTART to be available to students across the country within six years.
Top
Why2000: A Tutor that Teaches Mental Models Using Natural Language Dialogs
Funded by: Office of Naval Research
Funding period: 2000-2005
Amount: $1,168,700 for University of Memphis (Art Graesser, PI; Kurt VanLehn, PI at University of Pittsburgh), additional $90,175 for graduate student fellowship
Summary: Studies indicate that human tutors provide the most effective form of instruction known (Bloom, 1984; Cohen, Kulik, & Kulik, 1982). They raise the mean performance about two standard deviations compared to students taught in classrooms. Intelligent tutoring systems offer excellent instruction, but not quite as good as human tutors. The best ones raise performance about one standard deviation above classroom instruction (e.g., Anderson, Corbett, Koedinger, & Pelletier, 1995). In other words, a human tutor can raise the student's grade by about two letter grades (e.g., from C to A) while a tutoring system can raise it by about one letter grade (e.g., from C to B). Our challenge is to create tutoring systems that are as effective as human tutors.
The biggest remaining difference between human tutors and tutoring systems is that human tutors use natural language and most tutoring systems do not. Although tutoring systems print natural language and sometimes speak it, they do not let the student type or speak in natural language. They accept only constrained language such as menu selections, mathematical expressions, key words, etc. This limits students' opportunities to generate deep explanations and have the tutor critique them. Since student generation of explanations is known to increase learning, we hypothesize that if tutoring systems could participate in natural language dialogs with students about deep explanations, then the systems would become much more effective, and possibly rival human tutors. This could have a significant impact on military, commercial and public education.
This research project will (1) develop tools for building tutoring systems that conduct explanation-based natural language dialogs, (2) use the tools to develop tutoring systems for at least two task domains, and (3) evaluate their effectiveness compared to expert human tutors and to versions of the systems that use constrained language instead of natural language.
Our basic approach is to combine a shallow, statistical approach (Latent Semantic Analysis) with deep, symbolic approaches (the LCFlex parser, the Tacitus-Lite+ discourse interpreter, and the APE tutorial dialog manager). Although prototypes of these components have been developed in our earlier work, all require significant extensions to handle explanation-based tutorial dialogs. Such dialogs are less constrained than dialogs handled previously by symbolic approaches, and they require deeper processing than dialogs handled previously by statistical approaches. We believe our hybrid approach will yield both the robustness and depth of understanding that explanation-based tutorial dialogs require.
Top
Developing AutoTutor for Computer Literacy and Physics
Funding agency: National Science Foundation
Funding period: 2001-2004
Amount: $1,274,075 (Art Graesser, PI)
Summary: The Tutoring Research Group at the University of Memphis has developed a computer tutor (called AutoTutor) that simulates the discourse patterns and pedagogical strategies of unaccomplished human tutors. The typical tutor in a school system is unaccomplished in the sense that the tutor has had no training in tutoring strategies and has only introductory-to-intermediate knowledge about the topic. The development of AutoTutor was funded by an NSF grant. The discourse patterns and pedagogical strategies in AutoTutor were based on a previous project that dissected 100 hours of naturalistic tutoring sessions.
AutoTutor is currently targeted for college students in introductory computer literacy courses, who learn the fundamentals of hardware, operating systems, and the Internet. Instead of merely being an information delivery system, AutoTutor serves as a discourse prosthesis or collaborative scaffold that assists the student in actively constructing knowledge. AutoTutor presents questions and problems from a curriculum script, attempts to comprehend learner contributions that are entered by keyboard, answers student questions, formulates dialog moves that are sensitive to the learner's contributions (such as short feedback, pumps, prompts, assertions, corrections, and hints), and delivers the dialog moves with a talking head. The talking head displays emotions, produces synthesized speech with discourse-sensitive intonation, and points to entities on graphical displays. AutoTutor has seven modules: a curriculum script, language extraction, speech act classification, latent semantic analysis (a statistical representation of domain knowledge), topic selection, dialog management, and a talking head. Evaluations of AutoTutor have shown that the tutoring system improves learning with an effect size that is comparable to typical human tutors in school systems, but not as high as accomplished human tutors and intelligent tutoring systems. The dialog moves of AutoTutor blend in the discourse context very smoothly because students cannot distinguish whether a speech act was generated by AutoTutor or a human tutor.
This research will substantially expand the capabilities of AutoTutor by designing the discourse to handle more sophisticated tutoring mechanisms. These mechanisms should further enhance the active construction of knowledge. One enhancement is to get the student to articulate more knowledge, with more formal, symbolic, and precise specification; if the student doesn't say it, it is not considered covered by AutoTutor. Another enhancement is to set up the dialog so that it guides the user in manipulating a 3-dimensional microworld of a physical system; the student attempts to simulate a new state in the physical system by manipulating parameters, inputs, and formulae. The research will develop AutoTutor in the domains of both computer literacy and Newtonian physics, so we will have some foundation for evaluating the generality of AutoTutor's mechanisms. AutoTutor has been designed to be generic, rather than domain-specific; an authoring tool will be developed that makes it easy for instructors to prepare new material on new topics. After the new versions of AutoTutor are completed, we will evaluate its effectiveness on learning gains, conversational smoothness, and pedagogical quality. During the course of achieving these engineering and educational objectives, the project will conduct basic research in cognitive psychology, discourse processes, computer science, and computational linguistics.
Top
Understanding in Science
Funding agency: National Science Foundation, Understanding in Science
Funding period: 2002-2004
Amount: $296,902 for University of Memphis subcontract (Art Graesser, PI; Jennifer Wiley, PI at University of Illinois at Chicago)
Summary: Scientific literacy in the electronic age requires the use and coordination of multiple forms of media in a complex learning environment. Resources on the web and other computer faculties include electronic text, graphs, tables, pictures, animated simulation of mechanisms, and other forms of input. In the best of worlds, the various forms of media are strategically accessed and used when cognitive tasks are performed. Scientific literacy depends on an effective coordination of electronic multimedia, so it is imperative that we understand the cognitive mechanisms that are recruited when the various media are used.
The tasks of interest in this research require a deep understanding of the material rather than the mere accumulation of shallow facts. In particular, we will investigate tasks that require (a) multiple sources of information, (b) deep reasoning about scientific and technical content, and (c) the writing of multi-sentence compositions. Examples of such tasks are formulating research questions, answering deep-reasoning questions (why, how, what-if, what-if-not, what-are the consequences), constructing justified arguments, and evaluating the quality of evidence for a claim. These tasks are central to the practice of science, but are extremely challenging and are not adequately supported in traditional learning environments.
At this point, there has been virtually no research on the use of electronic media and multiple information sources when adults perform writing tasks that require deep reasoning. Fortunately, we can turn to some theories and models of cognitive processing that offer some guidance in our investigations of this uncharted territory. The proposed research will test some theoretical models of question asking, question answering, argumentation, comprehension, writing, and multimedia processing that have evolved in the fields of cognitive psychology, discourse processing, and education. We will test these theoretical models by collecting eye tracking data and/or think aloud data while college students view electronic multimedia and prepare written compositions that are required for particular pedagogical tasks. We will experimentally manipulate both the media available to learners and the tasks to be performed. Auxiliary measures will also be collected that tap memory, learning, and individual differences. The patterns of eye tracking data, cognitive measures, and behavioral measures reflect the cognitive strategies that underlie the use of electronic multimedia.
Top
Developing and Testing a Computer Tool that Critiques Survey Questions
Funding agency: National Science Foundation
Funding period: 2000-2003
Amount: $205,990 (Art Graesser, PI)
Summary: This research will further develop and test a computer tool that assists designers of questionnaires to improve the comprehensibility of questions. The computer tool has particular modules that diagnose each question in a survey on various levels of language, discourse, and world knowledge. For example, the critique identifies words that are unfamiliar to most respondents, ambiguous terms, questions that have complex syntax, questions that overload working memory, questions with misleading presuppositions, and questions that may appear to the respondents to be unrelated to the survey context. Some of these components are so complex, technical, or subtle that they are invisible to the unassisted human eye, including experts in survey methodology, questionnaire design, and computational linguistics. The computer tool integrates empirical findings and computational architectures in fields of cognitive science, artificial intelligence, computational linguistics, discourse processing, and psychology. The computer aid can be used as an automated tool to the extent that it provides an accurate, reliable, and fully automated diagnosis of problematic questions. Alternatively, survey methodologists can also use the tool to assist them in improving questions and to learn about the mechanisms that underlie the tool.
A first version of the computer tool was developed in 1999 in a research collaboration between the Institute for Intelligent Systems at the University of Memphis (Dr. Graesser is PI on contract) and the US Bureau of Census (with Dr. Marquis). The computer tool is called QUAID, which stands for the QUEST Questionnaire Evaluation Aid. QUEST is a cognitive computational model of human question answering that was previously developed by Dr. Graesser. Therefore, the QUAID is grounded in models of human cognition in addition to incorporating contemporary developments in computational linguistics (such as lexicons, syntactic parsers, neural network classifiers, and latent semantic analysis). We are currently evaluating the performance of QUAID with respect to 9 problems with questions that previous studies have found to be common on factual surveys, such as the US Census. One index of performance is whether the tool is discriminating in identifying problems with questions that are identified by experts in survey methodology and language. Other indices of performance address the difficulty that respondents have in comprehending the meaning of the questions.
The research will attempt to improve the 9 QUAID modules and will evaluate the revised QUAID tool on several performance measures. Each module determines whether or not a particular question has a problem (e.g., unfamiliar technical term, working memory overload). These decisions will be compared with the decisions of experts. Other performance measures are needed because trained expert judges may miss subtle computational mechanisms. These other measures assess whether QUAID's output can predict the behavior of respondents when they answer the questions: (a) behaviors of respondents that indicate they are having difficulty comprehending the question in a conversational interview (such as clarification questions of respondents) and (b) test-retest reliability of answers to questions when respondents answer a question on multiple occasions. Performance measures will also be collected for original questions, questions revised by survey methodologists who do not use of the QUAID tool, and questions revised by survey methodologists who use the QUAID tool.
Top
Monitoring Emotions While Students Learn with AutoTutor
Funding agency: National Science Foundation
Funding period: 2003-2008
Amount: $1,250,000 (Art Graesser, PI)
Summary: The intellectual merit of the proposed research is to build and test learning environments that coordinate complex learning and learner emotions. The project augments an existing intelligent tutoring system (AutoTutor) that helps learners construct explanations by interacting with them in natural language and helping them use simulation environments. AutoTutor has an animated conversation agent and a dialog management facility that attempts to comprehend the learner's contributions and to responding with appropriate dialog moves (such as short feedback, pumps, hints, prompts for information, assertions, answers to student questions, suggestions for actions, summaries). AutoTutor has been developed for computer literacy and qualitative physics; it is available on the web and on desktop applications. The tutorial dialogue of AutoTutor will be enhanced in the proposed research by incorporating signal processing algorithms and sensing devices that classify various facial patterns and affective states of learners. The proposed research has three specific objectives: (1) To assess patterns of facial activity that arise while interacting with AutoTutor and to identify pedagogically significant subsets of these patterns, (2) to investigate whether learning gains and learner's impressions of AutoTutor are influenced by dialog moves of AutoTutor that are sensitive to the learner's affective states, and (3) to test and augment theories that systematically integrate learning and emotion into educational technologies.
This proposed research investigates strategies, processes, practices, and environments that are likely to assist the learner in the active construction of knowledge, particularly at deeper levels of comprehension and problem solving. It is already well documented that knowledge construction is enhanced by one-on-one tutoring activities that encourage the students to articulate explanations of complex systems, as well as manipulating such systems in simulation environments. Simply put, students learn by telling and doing. These constructivist principles have already been implemented in AutoTutor and have been shown to produce robust learning gains. However, constructivism is not entirely limited to cognition, discourse, action, and the environment because emotions are inextricably bound to the learning process. An agile learning environment that is sensitive to a learner's affective state presumably enriches learning, particular when deep learning is often accompanied by confusion, frustration, boredom, interest, excitement, and insight. Therefore, attempts will be made to integrate state-of-the-art affect-sensing technology with AutoTutor. Emotions could potentially be classified on the basis of facial expressions, gross body language, spoken voice stress analysis, haptic sensing, galvanic skin response, and heart rate, but the primary focus in this research will be on facial expressions (together with knowledge states and discourse patterns). In addition to advancing research on complex learning, emotion, and sensing technologies, the proposed research will advance theories and models in the fields of cognitive science, discourse processing, computational linguistics, artificial intelligence, data mining, distance learning, and information technologies.
The broader impacts of the proposed research activities are to advance education, intelligent learning environments, and human-computer interfaces. It is widely acknowledged in the field of education that students rarely acquire a deep understanding of the material they are supposed to learn in their courses. Students normally settle for shallow knowledge, such as lists of concepts, a handful of facts about each concept, and disconnected definitions of key terms. Students lack the deep coherent explanations that organize the shallow knowledge and that fortify the learner for generating inferences, solving problems, and applying their knowledge to practical situations. Yet model-based reasoning is essential for learners of science, technology, engineering, mathematics, and other disciplines that are in high demand in today's workforce. The proposed research will attempt to fortify future learners with enhanced dynamic reasoning, automated cognitive assessment, and intelligent handling of emotions. A learning environment that monitors learner emotions is also likely to be more motivating and personally relevant to the learner.
Top
A Web Site on Institutional Review Boards and the Ethical Use of Human Subjects
Funding agencies: Institute for Defense Analysis & ONR NAWCTSD
Funding period: 2000-2001
Amount: $950,000 (Art Graesser, PI; $750,000 through Thoughtware Corporation and $200,000 through The University of Memphis)
Summary: The Human Use Regulatory Affairs Advisor (HURAA) was developed as a web-based facility that provides help, training, and information retrieval on the ethical use of human subjects in research. The content of HURAA is derived from Federal agency documents and regulations, particularly the National Institutes of Health (NIH, 45 CFR 46), the Department of Defense (32 CFR 219; DoD Directive 3216.2), and doctrine from particular branches of the US military. The targeted users of the HURA Advisor are high-ranking military officials who must review or approve research protocols involving human subjects in studies sponsored by the Department of Defense (DoD). These users focus on fundamental ethical issues, but not the detailed procedures and paper work associated with gaining approval from Institutional Review Boards.
The design of HURAA was guided by a number of broader objectives. First, the layout and design of the web facility was to incorporate particular guidelines in human factors, human-computer interaction, and cognitive science. Second, the architecture of the HURAA components needed to be conformant with the ADL standards for reusable instructional objects, as specified in the Sharable Content Objects Reference Model (SCORM, 2001). The primary objective of having these standards is to allow course content to be shared among different lesson planners, computer platforms, and institutions. Third, HURAA was designed to optimize both learning and information transmission. High-ranking military personnel have very little time, so it was imperative to promote the speed and quality of learning in web-based distance learning environments. This required careful consideration of the pacing of the information delivery, the selection of content, and design of the tasks to be performed. One version of HURAA, for example, included an animated conversational agent that served as a navigational guide for the web facility. Fourth, the web site was supposed to be engaging to the user. Otherwise, high-ranking officers would lose interest before completing the training. HURAA included some persuasive multimedia that was intended to hook the user to continue using the website.
A fifth broader objective behind the design of HURAA was to incorporate recent learning technologies. HURAA incorporated some of the sophisticated pedagogical techniques that have been implemented in advanced learning environments with intelligent tutoring systems and animated conversational agents. HURAA has a number of standard features of conventional web facilities and computer-based training, such as hypertext, multimedia, help modules, glossaries, archives, links to other sites, and page-turning didactic instruction. HURAA also has more intelligent features that are hypothesized to promote deeper mastery of the material, such as lessons with case-based and explanation-based reasoning, document retrieval though natural language queries, animated conversational agents, and context-sensitive Frequently Asked Questions called "Point & Query". Three experiments were conducted to test the users' perceptions, usage, and learning from different versions of HURAA that either contained or did not contain these advanced features.
A sixth broader objective behind the HURAA design was quick access to relevant information. HURAA attempted to optimize information retrieval when users actively search for information. Some users prefer to learn by actively exploring information, asking questions, and accessing answers to questions, whereas other learners tend to be more passive and rarely pose questions. A good web facility provides information quickly on demand for those occasions when users are curious, active, self-regulated learners. HURAA implemented multiple methods of accessing information in a large space of documents. HURAA has hypertext and glossaries, which are conventional methods of information retrieval. HURAA also has (a) a Point & Query facility, which provides answers to context-sensitive Frequently Asked Questions (FAQs), and (b) a facility for accessing documents that provide answers to questions that the user formulates in English, using techniques in computational linguistics. Three experiments were conducted to evaluate the extent to which the documents fetched by natural language queries yielded answers that were relevant and informative.
The design, development and testing of HURAA was an evolutionary process of rapid prototyping and testing, not a process of design from specification. This was necessary because: (a) ADL/SCORM and the new learning technologies were rapidly changing in fundamental ways, (b) the requirements in handling research ethics through this form of instructional medium in the DoD were evolving, HURAA being the first instance, and (c) the content and design of HURAA required feedback from multiple sectors of the DoD. The process of rapid prototyping and testing has the advantages of producing quick initial products and of being responsive to a dynamic R&D environment. However, the down side is that the process is not well disciplined. Therefore, one of the missions of this project was to document the history of designing, developing, testing, and documenting HURAA. This history includes a critical self-examination of the successes and failures of the design and evaluation process. The lessons learned from this project will hopefully be beneficial to the entire ADL community.
Top
Promoting active reading strategies to improve students' understanding of science
Funding agency: National Science Foundation IERI
Funding period: 2000-2005
Amount: $3,196,000 (Danielle McNamara, PI)
Summary: In recent years, there has been a growing concern that high school students in the United States have been falling in terms of their reading ability and science knowledge. Project Strategies is a 5-year National Science Foundation grant designed to address high school students' comprehension in their science courses. The project incorporates a team of approximately 15 scientists from multidisciplinary fields including cognitive psychology, computer science, linguistics, school psychology, and discourse processes.
There is growing concern with regards to the current state of the educational system the United States. For example, in comparison to other countries, high school students in the United States are falling behind students in other countries on various measures of academic achievement, and in particular, on measures of reading comprehension (National Assessment of Educational Progress, 1998; Snow, 2002). To compound the problem teachers rarely provide instruction on strategies that emphasize comprehension (e.g., Durkin 1978-79; Taylor, Pearson, Clark, & Walpole, 1999), and students rarely engage in such strategic processing (Pressley & Ghatala, 1990; Rothkopf, 1988; Garner, 1990; Pressely et al., 1992). In many cases, students resort to primitive and ineffective methods of learning such as repetition by default (Garner, 1990). Such mindless processing not only leads to a poor memory of an event, but also leads to a shallow representation and understanding of the domain. The corollary of shallow processing is poor performance on tasks that require the transfer or application of acquired knowledge (Gick & Holyoak, 1980; Reeves & Weisberg, 1994).
To compound the problem, many high school texts are not written in manner conducive to deep comprehension. Many high school texts are written in a format that is "inconsiderate" to new learners (Beck, McKeown & Gromoll, 1989). Textbook authors often leave out background information because they assume (inappropriately) that readers are already aware of the surrounding issues. Lack of prior knowledge has been shown to be detrimental for text comprehension (e.g., Bransford & Johnson, 1972). Moreover, when new topics are discussed, they are often covered at such a broad and general level that students cannot develop a deep-level of understanding. To make matters worse, textbook material is often presented as a set of unconnected statements, leaving it up to the reader to draw the proper inferences and to build the appropriate relations between the concepts. Unfortunately, many readers do not normally engage in such strategic processing (Pressley & Ghatala, 1990; Rothkopf, 1988), and the end result is poor comprehension and academic difficulties. In sum, much more research is needed to both explore the reasons why students are not achieving at optimal levels, and to develop, evaluate and implement solutions for improving pedagogy.
The project has three major phases. Phase one includes the assessment of student abilities including reading ability, knowledge, strategy use, demographics, student activities and habits. The second phase concerns the design and evaluation of various learning strategies intended to improve students' achievement by targeting how students learn new information and solve problems. The third phase of the project is geared towards the development of an automated computer tutor called iSTART. The tutor bridges the information gained in phases one and two to produce a strategy trainer that can be incorporated in the classroom. iSTART teaches students to be more effective readers and thinkers; the benefits of which are expected to generalize beyond the classroom.
Top
Coh-Metrix: Automated Cohesion and Coherence Scores to Predict Text Readability and Facilitate Comprehension
Funding agency: Institute of Education Sciences
Funding period: 2002 -2005
Amount: $1,425,200 (Danielle McNamara, PI)
Summary; Text cohesion is a driving factor of comprehension and text characteristics critically interact with readers' abilities. By using advanced computational linguistic tools thoroughly tested in experimental research, this project provides the latest in experimental psychology, computational linguistics, education and computer science to measure and improve readability of texts.
The last three decades of reading comprehension research has led to a more complete understanding of reading comprehension processes. This evolution of reading theory, in conjunction with recent developments in discourse processes and computational linguistics has led us to a point where traditional readability formulas (e.g. Flesch-Kincaid scores) can be dramatically improved.
Our primary challenge in the proposed grant is to develop an automated cohesion metric tool (Coh-Metrix) that computes properties of text cohesion and that computes a coherence score that integrates text cohesiveness with the reader's world knowledge and aptitude. We will also develop an automated tool that identifies specific cohesion gaps in text (Coh-GIT).
Our second goal is to further our understanding of the complex interactions among the text, reader, tasks, and levels of representation. This will allow us to better calibrate Coh-Metrix and Coh-GIT in addition to advancing theories of reading comprehension. We plan on conducting several experiments to investigate effects of text cohesion for young readers (grades 3-5) and young-adult readers, and how those effects depend on text genre, prior knowledge and word-based reading skills
Our third goal is to fine-tune and validate our coherence measurements by (a) using existing data corpora and data collected during this project to calibrate and test our metrics; (b) examining and comparing cohesion metrics and readability scores across a large set of K-14 basal readers and instructional textbooks, (c) empirically verifying the validity of Coh-GIT by identifying cohesion gaps in texts and comparing it to eye-tracking patterns.
The potential applied contributions of the Coh-Metrix and Coh-GIT tools are innumerable. These tools will allow readers, writers, editors, and educators to more accurately estimate the appropriateness of a text for their audience, and pinpoint specific problems with text. Thus, this project will be of benefit to both practitioners and policy makers, helping to improve students' ability to comprehend and learn from text.
Top
Data-Link Aircraft Communications: An Examination of Interference, Coherence, and Workload
Funding agency: National Aeronautics and Space Administration AMES Research Center
Funding period: 2001-2004
Amount: $365,000 (Danielle McNamara, PI)
Summary: Recent technology allows Air traffic control (ATC) messages to be sent in textual format via data link, as well as via radio transmission. The purpose of this project is to investigate differences between auditory and textual transmission of information to better predict errors as a function of modality and to ultimately develop a set of guidelines for data link transmission of ATC messages. To this end, we will address three issues: working memory interference, message coherence, and workload.
The first issue specifically regards whether different modes of interference (verbal or visual) are more detrimental for data link or radio transmission. We will also investigate differences between interference that occurs concurrently and subsequent to the target task.
The second issue regards the effects of coherence of the ATC message.We will compare auditory and textual presentation of data-link messages that vary in coherence on two levels, the word (spelled out vs. abbreviated) and the idea (complete, incomplete).
The third issue regards the differential effects of workload on text and speech communications. Moderate workload will involve performing the data-link activities simultaneously with the monitoring task. High workload will involve data-link activities performed simultaneously with monitoring, tracking, and resource management tasks. A high-workload/unpredictable condition will additionally include unexpected failures in the resource management task. Within each of these areas we also investigate effects of message length, urgency, and expectancy. This project will provide valuable information concerning the effects of interference, message coherence, and workload on task completion and comprehension in the ATC environment.
Top
Percolation Models of Phase Transition during Perceptual Processing
Funding agency: National Science Foundation /Quantum & Biologically Inspired Computing
Funding period; 2002-2005
Amount: $652,194 (Robert Kozma, PI)
Summary: The Neuropercolation project is a 3 year grant by the National Science Foundation on "Dynamical Behavior in Percolation Models Related to Phase Transitions in the Cortex during Sensory Information processing." The Neuropercolation project is funded by the NSF Directorate for Experimental and Integrative Activities, in the framework of the Biological Information Technology Systems Program (BITS), also related to Quantum and Biologically Inspired Computing Program (QuBIC). The period of this grant is from March 1, 2002 to February 28, 2005. These web pages will contain a complete and ongoing record of the artifacts, publications and other products that result from the research supported by this grant.
Top
Self Organizing Ontogenetic Development for Autonomous Adaptive Systems
Funding agency: NASA/Ames Revolutionary Computing Systems
Funding period:
Amount: Robert Kozma, PI
Summary: The SODAS project is a 3 year grant for the exploration and development of models of self-organizing representations and behavior in biological and artificial organisms. Our work combines insights from neuroethology, artificial intelligence and cognitive science. In particular, we are interested in how the organization of structure and behavior in biological organisms can be viewed as a self-organizing dynamical system. We are actively exploring the role that chaotic modes of dynamics may play in this self-organization of behavior.
The SODAS project is funded by the NASA group for Research in Intelligent Systems. The period of this grant is from January 1, 2001 to December 31, 2003. These web pages will contain a complete and ongoing record of the artifacts, publications and other products that result from the research supported by this grant.
Top
Immunity-Based Intrusion Detection Systems
Funded agency: DARPA
Funding period: 2000- 2003
Amount: Dipankar Dasgupta, PI
Summary: The goal of this project is to develop an intelligent multi-agent system for intrusion/anomaly detection and response in networked computers. The approach is inspired by the defense mechanisms of the immune system that is a highly distributed in nature. In this approach, immunity-based agents roam around the machines (nodes or routers), and monitor the situation in the network (i.e. look for changes such as malfunctions, faults, abnormalities, misuse, deviations, intrusions, etc.). These agents can mutually recognize each other's activities and can take appropriate actions according to the underlying security policies. Specifically, their activities are coordinated in a hierarchical fashion while sensing, communicating and generating responses. Moreover, such an agent can learn and adapt to its environment dynamically and can detect both known and unknown intrusions.
Top
Preliminary Research on Immunity-Based Computational Techniques
Funding agency: NSF SGER
Funding period: 2001-2002
Amount: Dipankar Dasgupta, PI
Summary: The PI will conduct a preliminary investigation of immunity-based computational techniques to pave the way for more complex studies of this subject in the future. The ultimate goal of this research is to develop computational techniques inspired by the natural immune system for solving real-world science and engineering problems. The natural immune system is a distributed novel-pattern recognizer, which uses intelligent mechanisms to detect a wide variety of antigens (novel patterns). From the computational point of view the immune system uses learning, memory, and associative retrieval to solve recognition and classification tasks. The immune system is a subject of great research interest, not only in the hope of finding cures for many diseases but also as a means for understanding its powerful information processing capabilities. In the current project the PI will investigate immunological principles, explore the underlying concepts and mechanisms, and take initial steps towards the development of intelligent computational techniques for solving problems in the field of science and engineering.
|
