David Kirsh - US grants

The purpose of this workshop on the foundations of Artificial Intelligence is to systematically explore the scope and limits of fundamental approaches to artificial intelligence. In contrast to previous efforts, this workshop is structured to generate genuine discussion and interaction between paradigms. The workshop will have eight sessions in which a careful exposition of each approach will be presented by a founder of that approach, and will then be critiqued by a respected and influential commentator. Each position paper contains two parts. In the first part, the founder states what the basic thesis is, identifies the principles underpinning the method or approach and describes the natural type of problems and tasks in which this approach succeeds. In the second part, the speaker discusses the scope and limits of the approach, identifying particular difficulties and providing a concrete problem that illustrates each. The commentator then critiques the scope and limits of the approach, where it succeeds and why, and where it fails and why. The NSF funds will be used to provide support for conducting the workshop as well as widely disseminating the written materials presented at the workshop and videotapes of the presentation and ensuing discussions.

Dr. Robert Cummins of the University of Arizona and Dr. David Kirsh of the University of California, San Diego, are analyzing and comparing two radically different approaches to the study of cognition and intelligence, the orthodox "computational" approach, on the one hand, and the "new connectionist" or parallel distributed processing approach on the other. The orthodox computational approach takes mental processes to be analogous to the processes in a programmed digital computer. Drs. Cummins and Kirsh are undertaking a systematic analysis and comparison of these paradigms in a scientifically and philosophically informed way. The goals of this project are 1) to compare the research methods and explanatory goals of the connectionist and computational approach to cognitive science; 2) to articulate the underlying conception of cognition, intelligence and mentality that drives exponents of each approach; and 3) to assess the possible impact on cognitive science of the major shift of resources into connectionist research that is currently taking place. The investigators will carry out site visits to major centers of connectionist and orthodox computational research. They are targeting two specific research areas for investigation on these visits: speech perception ( a strength of connectionism) and problem solving (a strength of the orthodox approach).

This is a three-year continuing award. To ensure that new workplaces and digital work materials serve human needs requires both a different theoretical base and an integration of theory with ethnographic and experimental methods. We propose a research framework based on distributed cognition. Distributed cognition provides a reorientation of how we think about designing and supporting human computer interaction. It permits one to move boundary of the cognitive unit of analysis out beyond the skin of the indivdual to include the material and social environment as components of a larger cognitive system. This focuses attention on the processes by which people take advantage of both internal and external resources to organize their actions. For the design of workplaces, this means that work materials are more than stimuli for disembodied cognitive sytems. Work materials become elements of the cognitive system itself, and cognition becomes an emergent property of the interactions among people and work materials. We will apply our integrated approach to developing a distributed cognition based theory of annotation and explore a range of application domains: collaborative scientific research, education, and commercial aviation. In each domain, we will conduct ethnographic and experimental studies, design and implement digital work materials and analyze their effectiveness. In addition, to help us develop and communicate our distributed cognition perspective, we are developing: (1) a series of prototypes of history-enriched work materials that support collection and selective sharing of personnel and group activity histories and annotations, (2) annotation facilities to assist collaborations, and (3) information visualization facilities to exploit rich histories of interactions and provide effective access to annotations.

People have considerable experience using 3D graphics in scientific visualization, but very little understanding of exactly what aspects of those 3D graphics give us the most perceptual benefit. As a consequence, we do not understand how to optimize and improve our use of this tool. In this project, the PIs will perform human-subject tests to test the effectiveness of various depth cues on the perception of 3D scenes, and to quantify the difference in effectiveness between perceptual depth cues in synthetic scenes and in "real" scenes as presented by physical models. The findings will have significant implications for interactive computer graphics and scientific visualization in general, and for virtual reality in particular, where much effort and money are being invested to generate realistic synthetic scenes without any rigorous scientific knowledge of what the most convincing realistic-scene cues really are. This project should supply some of the answers.

[unreadable] DESCRIPTION (provided by applicant): [unreadable] [unreadable] On any given weekday, more than one-fifth of the U.S. population is in schools. While we all want schools to be safe places-havens for learning-schoolchildren today face serious health threats including bioterrorism. In the proper project, we will develop and evaluate a system called ASIAN (Absenteeism-causing Symptom-Level Assessment Network), designed to monitor the health of children in schools. Our system focuses on monitoring children's school attendance, because most experts agree that absenteeism from school is an early sign of an outbreak in a community. However, school absenteeism is not a precise sign of an outbreak or epidemic. Absenteeism is often not even related to illness, reflecting, instead, problems in the child's home environment, important social or cultural obligations, and even opportunities for family recreation. But what if a system could identify absences due to illnesses and collect data on the symptoms causing the illness-related absences? Such a system might well offer an early, highly specific signal of an illness in a population, allowing public and school health officials to detect outbreaks in schools very early in their course. We propose to develop a VoiceXML speech processing system that can accurately capture during a brief telephone interview the symptoms a person is experiencing that are causing an absence. We propose to modify the "work processes" school districts use to have parents or guardians report that their children will be absent from school. Parents will use the interactive voice response questionnaire (or a web site, if they prefer) to report (on a voluntary basis) the symptoms that are causing the child's absence. We believe that monitoring of reports of such symptoms could both protect the health of children from bioterrorist attacks and from toxins in the environment. It could also improve their health on a daily basis by helping school nurses identify children with poorly managed chronic illnesses, and recognize and manage the routine outbreaks of infectious diseases that occur each year. The study plan focuses on creation of a computer telephone interviewing system and a formative evaluation that will include both qualitative and quantitative components and will assess the feasibility, validity, and potential usefulness of the system. [unreadable] [unreadable] [unreadable] [unreadable]

This SGER project explores a new area of study in which little is known: how skills learned within virtual environments (VE's) such as Nintendo's Wii are transfered to real-life natural environments. Part of what makes this study high-risk is that there many factors to control for to systematically compare skill transfer from and to virtual and real-life environments. The goal of this exploratory project is to investigate the mechanisms transfer through three mixed-methods experiments involving cooking. In this project new ethnographic methods for encoding what and how skills are acquired in VE's will be developed as appropriate (e.g., through video analysis). The PI is a world leader in "distributed cognition" and is the ideal person to lead this preliminary investigation of skill acquisition across real and virtual environments.

The broader impact of this research is in its potential to uncover new affordances of virtual environments for task acquisition, with implications for VE design. Through validating ethnographic methods for studying virtual to real-world transfer, this project will benefit other researchers. Further, these experiments serve as an important first contribution toward a future body of research that examinines how combining virtual and real-world training/instructions may best optimize human performance.

The intellectual history of science and computer science has placed undue weight on the propositional, symbolic and linguistic side of thought; leaving unexplained how thought can occur in non-verbal systems, based closer to our senses, and often coordinated with non-verbal processes in the socio-technical world. The goal of this study is to develop a new computational model of creativity based on an experimental and ethnographic study of choreographic invention. The core question is how the distributed system consisting of choreographer and dancers are able to be so generative. In all design and research work, creativity inevitably makes important use of non-propositional thinking. It is important, therefore, to deepen our insight in this ill-understood process. A further reason to study choreographic production is that it resembles large design and research projects in taking place over many weeks or months, it involves the collaboration of multiple parties, and a new production may easily cost a million dollars. At a group level the dynamics are similar to other large design efforts.The choreographer being studied has developed techniques for keeping the generative phase of the creative process open longer and for maintaining substantial variance among the dancers despite the urge for group think and convergent behavior. He has also developed techniques for exploiting the coding language of sensory systems, of both himself and his dancers, to create new movement ideas.

In this project shareable ethnographic and experimental data will be gathered that broadens the development of new computational theories about: 1) Distributed creativity: how the distributed components work to generate creative product - the mechanisms by which team members harness resources to interactively invent new concepts and elements, and then structure things into a coherent product; and 2) Embodied cognition: how the embodied aspect of cognition is harnessed to generate new 'thought' - the mechanisms by which designers, engineers, artists, dancers, and scientists think non-propositionally, using parts of their own sensory systems as simulation systems, and in the case of dance, using their own (and other's) bodies as active tools for physical sketching. The close study of both of these processes bears directly on the goals of developing new theoretical models of creativity and new models for research and education that will foster and reward creativity. The theory relocates creativity from a within-the-mind process to a more socio-technical process involving resources and other people; and it recognizes the important role that bodies and sensori-motor systems - both non-verbal and perhaps sub-rational elements - play in creative cognition.

The project proposes to create a distributed, multi-user social computing environment that will develop the capabilities of human Electroencephalography (EEG) to analyze users engagement with digitally based experiences. For this project, users will wear non-invasive, EEG headsets while navigating a shared virtual world. Beginning with a handful of EEG systems, the team will scale up over the course of the project to gather signals from dozens of users, providing a basis for larger scale studies. By comparing the EEG signals with each participants activities in virtual world, and with the brain activity and the activities of other users, a model of human brain activity will be developed for different types of behavior profiles and subjective states. This will allow significant improvement for the development of neural markers of human perceptual, cognitive and affective states, the parsing of EEG signals, the applicability of EEG interfaces to new types of experiences, all of which can enhance distance learning, collaborative distributed work, improved mobile computing interfaces and health care applications. The project will advance the capabilities for determining an individuals cognitive state by the creation of new computing methods utilizing comparative EEG analysis and data analysis of event states in a digital simulation. Bringing methods of large scale data analysis to articulate patterns across many users in the situated milieu of the online virtual world will create a new method to utilize EEG analysis to infer human subjective experience. The necessity of conducting this analysis in real-time, with data gathered from distributed, wireless EEG instruments will provide the impetus for utilizing accelerated hybrid multi-core techniques to bear on this domain.

The results from the project will be applicable for a variety of digital environments including computer aided learning and training, digitally mediated collaborative work environments, visualization of complex data sets, and digitally based entertainment experiences such as virtual worlds and computer games. The project will improve the functionality and outcomes of these digital media environments, better adjusting them to the cognitive states of the users. The PIs will train and employ a diverse body of participants to be involved in these activities. As part of an internship program at the Preuss School, a nationally recognized middle and high school on the campus of UCSD, UCSD faculty and staff mentor high school seniors to provide these students valuable experiences in a research laboratory. In addition, UMBC operates the very successful Meyerhoff program for minority students, primarily African-American, for which over 90% of a yearly entrance number of more than fifty go on to graduate school in science or engineering. Many of these students are computer science majors that have had the elective graduate course of Service Oriented Computing for Scientists and Engineers.