|
Home | People | Talks | Graduate | Undergraduate
Gary Lupyan, University of Pennsylvania Humans are the only animals to have names for their categories. Beyond making linguistic communication possible, do words (verbal labels) enable or facilitate certain cognitive processes? Does learning to name objects help us learn to what categories they belong? Does labeling a familiar object change how we remember it? Does hearing a word affect visual perception? I will talk about these issues in the context of the broad question—what do words do?—and present experimental evidence showing that words do far more than simply communicate information between individuals. Rather than viewing words as simple outputs of a conceptual system, I will argue that words (and language more broadly) should be viewed in terms of its feedback (top-down) effects on lower-level cognitive and perceptual processes. Dr. Gary Lupyan is a postdoctoral fellow at the University of Pennsylvania. He received his Ph.D. in Cognitive Psychology from Carnegie Mellon University and the Center for the Neural Basis of Cognition in 2007 under the advisorship of Jay McClelland. For the past year he has worked with Michael Spivey on the effects of language on visual processing. His interests include the role that language plays in nonverbal tasks (perception, categorization, memory), top-down effects on perception, semantics, and language evolution. Charles Elkan, UC San Diego Normally, the input for learning a binary classifier consists of positive training examples and separate negative training examples. However, the available training data often include an incomplete set of positive examples and a set of unlabeled examples, some of which are positive and some of which are negative. The problem solved here is how to learn a traditional binary classifier, given a nontraditional training set of this nature. Assuming that the labeled examples are selected randomly from the positive examples, I show that a classifier trained on positive and unlabeled examples predicts probabilities that differ by only a constant factor from the true conditional probabilities of being positive. Next, I show how to use this result in two different ways to learn a classifier from a non-traditional training set. Then, I apply these two new methods to solve a real-world problem: identifying protein records that should be included in an incomplete specialized molecular biology database. Experiments in this domain show that models trained using the new methods perform better than the previous state-of-the-art method for learning from positive and unlabeled examples. Dr. Charles Elkan received his Ph.D. in 1990 from Cornell University. He is currently a professor in the Department of Computer Science and Engineering at the University of California, San Diego. In 2005/06 he spent a sabbatical at MIT, and in 1998/99 he was Visiting Associate Professor at Harvard. Dr. Elkan is known for his research in machine learning, data mining, and computational biology. The MEME algorithm hedeveloped with his Ph.D. student, Tim Bailey, has been cited by over 1000 papers in biology and computer science. Dr. Elkan has won several best paper awards and data mining contests, and his Ph.D students have held tenure-track or equivalent positions at Columbia University, the University of Washington, the University of Queensland, other universities, and IBM Research. Michael A. Erickson, UC Riverside
It has been hypothesized that one learning approach in categorization tasks is the divide-and-conquer strategy (often referred to as "task partitioning"). This strategy offers benefits: Learners can limit their efforts to acquiring simple tasks rather than single complex one. These benefits, however, may come at a cost: Multiple tasks must be kept in mind, but not allowed to interfere with one another. Finding evidence of these costs would provide support for the use of task partitioning in category learning. In the research I present, the three relationships are examined: first, the relationship between task partitioning and task-switch costs; second, the relationship between executive attention and successful task partitioning.; and third, the relationship between task-switch costs and executive attention. I discuss the implications of these results for models of category learning. Michael Erickson is an Assistant Professor in the Department of Psychology at the University of California, Riverside. He received a B.S. Summa Cum Laude in Cognitive Science from the University of California, San Diego in 1993, and he received a Ph.D. in Cognitive Science and Psychology from Indiana University in 1998. He then completed post-doctoral training with Lynne M. Reder and James L. McClelland at Carnegie Mellon University. Following this training, he moved to the University of California, Riverside in 2002. His research interests include leaning, memory, and attention. His experimental methodologies include category learning, measures of working memory capacity, and negative priming, and his analytical tools include standard statistical procedures as well as computational and mathematical models of people's behavior. He received the 2004-2005 Professor of the Year award from the Department of Psychology, the 2005-2006 Teaching Award from the College of Humanities, Arts, and Social Sciences at UCR, and the 2005-2006 Innovative Teaching Award from the UCR Academy of Distinguished Teachers. Travis Seymour, UC Santa Cruz How do memory retrieval processes lead to overt responses in strategic recognition tasks (responding "old" to one class of familiar stimulus items and "new" to another)? Current theories of memory retrieval ignore the response requirements in such memory tasks and instead model them using memory processes (e.g., familiarity and recollection) alone. We argue that strategic recognition involves conflict in response processing similar to canonical conflict tasks (e.g., Stroop). This prediction has already been supported by symbolic computational modeling using the EPIC cognitive architecture. This prediction was further testing using brain activation and electromyographic measures. In one experiment we measured activation from anterior cingulate cortex, a brain region hypothesized to be sensitive to response conflict and found greater activity on trials where familiarity and recollection processes led to competing responses. An additional experiment used electromyography to replicate this finding with a measure incontrovertibly related to motor execution. Overall, results are consistent with the parallel task-set model (Seymour, 2001) assumption that recognition, motor, and control processes interact in strategic retrieval tasks. We conclude that modeling memory retrieval must include not only memory processes, but also careful consideration of any physical behaviors associated with retrieval. Dr. Travis Seymour is an Assistant Professor in the Department of Psychology at the University of California, Santa Cruz. He is a cognitive scientist who specializes in human memory and action. His research focuses on the cognitive limitations that sometimes constrain the use of optimal task strategies. These constraints can reveal important features of the cognitive architecture, and allow the prediction of behavior in new task settings. Dr. Seymour uses a variety of methodologies including behavioral, psychophysiological, and brain imaging approaches. The data from his studies are used to create dynamic computational simulations that not only serve as detailed theoretical models, but that can accurately predict human behavior in novel task environments. Alonzo C. Addison, UNESCO World Heritage Centre Just a few decades ago, computer graphic reconstructions of ancient monuments, laser scanning, and real-time GPS mapping of archaeological sites were unheard of, let alone immersive game engines and augmented reality. Now almost two decades old, the use of digital tools in cultural heritage, or "Virtual Heritage" has matured to the point where photorealistic reconstructions of past worlds are regularly on television, museums tout interactive immersive games, and anyone can construct 3D models of ancient structures with simple tourist photos via 3D webservices. Yet while digital recording, modeling, and dissemination of the past are
now commonplace, new challenges have arisen. With more and more of the world's heritage at risk from development to decay, and massive growth in the digital study of it, the need for a shared, coordinated, global Miguel Eckstein, UC Santa Barbara Image quality is often defined in terms of image properties such as contrast and resolution. However, in many applications such as in medicine images are typically visually inspected by humans. For these instances, image quality can be objectively defined in terms of performance of the human observer in the clinically relevant task (e.g., visually classifying a tumor as benign or malignant and/or localizing a nodule in the lung). In the first part of this talk I will discuss the importance of human perceptual performance in diagnostic decisions and the complexity of the various visual and cognitive processes used by radiologists. In the second part, I will describe some efforts in my laboratory to use computer models in conjunction with genetic algorithms to optimize medical image compression with respect to human observer task performance. Miguel Eckstein earned a BS in Physics and Psychology at UC Berkeley and a PhD in Cognitive Psychology at UCLA. He then worked at the Department of Medical Physics and Imaging, Cedars Sinai Medical Center and NASA Ames Research Center before moving to UCSB. He is recipient of the Optical Society of America Young Investigator Award, the Society for Optical Engineering (SPIE) Image Perception Cum Laude Award, Cedars Sinai Young Investigator Award, the National Science Foundation CAREER Award, and the National Academy of Sciences Troland Award. He has served as the chair of the Vision Technical Group of the Optical Society of America and the Human Performance, Image Perception and Technology Assessment conference of the SPIE Medical Imaging Annual Meeting, and as a member of the National Institute of Health study section panels. He currently serves as the Vision Editor of the Journal of the Optical Society of America A and the board of editors of Journal of Vision. He has published over 80 articles relating to computational human vision, visual attention, and the perception of medical images. Garrison W. Cottrell, UC San Diego There has been a great deal of progress in understanding how complex objects, in particular human faces, are processed by the visual system of the brain. Sophisticated neurocomputational models have been developed that do many of the same tasks accomplished by these cortical areas. Such models allow us to explore hypotheses concerning the relative contributions of nature versus nurture, the level of processing that is most responsible for explaining behavior in a task, and the nature of the features underlying face processing. In the last ten years, we have developed what is arguably one of the best models of face and object processing in terms of the amount of data we have been able to explain. These data include the interaction between expression and identity (or lack of it) in holistic processing, "categorical" and continuous data on facial expression processing, how the Fusiform Face Area (FFA) could become recruited for other expert-level processing tasks, and how specialized processing networks such as the FFA can arise from a combination of innate biases, task constraints, and competitive learning. In addition, we have used our model to explain human data in face and expert Chinese character processing, a stimulus set that shares a great deal with faces except for configurational effects, yet still shows the inversion effect. We have also investigated the role of hemispheric processing in expertise, showing how the left-side bias in face perception arises naturally from a model that has differential hemispheric spatial frequency processing. Unfortunately, our model is wrong in a fundamental way: it processes static images of faces and objects to the same level of detail everywhere in the image. That is, the model does not have a fovea, nor does it direct its gaze. In human vision, visual acuity falls off rapidly from fixation to periphery, and so we actively change our gaze direction to bring relevant information into foveal vision, where the highest-quality visual information can be obtained. In order to build such a model, one needs to model two aspects of saccade-based vision: first, where to look, and second, how to integrate information across saccades for recognition purposes. In the last few years, we have developed several Bayesian models in this domain: (1) a model of face and object recognition and classification that uses multiple fixations on objects and faces, rather than one "fixation" of the whole object or face as our previous model did; (2) a model of eye movements during concept learning; and (3) a model of visual salience in static images and in video. In addition, we have developed (4) an unsupervised, hierarchical model of independent feature extraction. This is important in probabilistic models, as independent features make probabilistic inference tractable. In this talk I describe some of these models and plans for future work. Dr. Garrison W. Cottrell received his Ph.D. in 1985 from the University of Rochester under James F. Allen. He conducted postdoctoral studies with David E. Rumelhart at the Institute of Cognitive Science at the University of California, San Diego until 1987, when he joined the faculty of the Computer Science and Engineering Department. He is currently Professor of Computer Science and Engineering and Director of the NSF sponsored Temporal Dynamics of Learning Center. Dr. Cottrell's main interest is Cognitive Science, specifically in building working models of cognitive processes and using them to explain psychological or neurological processes. In recent years, he has focused upon face processing, including face recognition, face identification, and facial expression recognition. He has also worked in the areas of modeling psycholinguistic processes, such as language acquisition, reading, and word sense disambiguation. Jay McClelland, Stanford This presentation will introduce a wide-ranging research effort involving primate neurophysiology, mathematical analysis, computational modeling, and human behavioral experimentation that examines how reward and stimulus influences are combined in real time to affect perceptual decisions. The human behavioral data indicates clear deviations from optimality, in different ways for different participants. Whether participants are striving for optimality in the face of differing limitations or let go of optimization because it is too hard to maintain are topics will be considered in discussion. Dr. Jay McClelland received his Ph.D. in Cognitive Psychology from the Thomas Griffiths, UC Berkeley Much of the knowledge that human beings have about their world comes not from direct experience, but from interacting with others. This raises an interesting question: what are the consequences of learners learning from other learners? I will present both theoretical and empirical results on the implications of such a process of "iterated learning" for the information being transmitted between learners. Specifically, I will show that iterated learning with Bayesian learners, each observing data generated by the previous learner and then selecting hypotheses in accordance with Bayes' rule, results in convergence to a distribution over hypotheses determined by the prior of the learners. This result has implications both for methods for developing methods for investigating the inductive biases of human learners, and for understanding how those biases influence the evolution of concepts and languages. I will describe the results of a series of experiments bearing out the basic predictions of analysis for simple cognitive abstractions such as functions and categories as well as more complex linguistic objects such as frequency distributions and systems of color terms. Dr. Tom Griffiths is an Assistant Professor in the Department of Psychology and the Cognitive Science Program at the University of California, Berkeley. He received his Ph.D. in Psychology from Stanford University in 2005, and taught at Brown University before coming to Berkeley. His research focuses on developing mathematical models of human cognition, with an emphasis on exploring rational accounts of behavior derived from computer science and statistics.
Several lines of research in psychology have led to the conclusion that Dr. Samuel M. McClure is Assistant Professor of Psychology at Stanford Jim Whitehead, UCSC Almost all software contains undiscovered bugs, ones that have not yet been exposed by testing or by users. Wouldn't it be nice if there was a way to know the location of these bugs? This talk presents two approaches for predicting the location of bugs. The bug cache contains 10% of the files in a software project. Through an analysis of the software's development history and the location of bugs, files are added and removed from the cache based on four notions of bug locality: temporal, spatial, changed-entity, and new-entity locality. After processing, files in the bug cache contain 73-95% of undiscovered bugs. To improve the localization of predicted bugs, the second prediction approach focuses on configuration management commit transactions. Using machine learning techniques (Support Vector Machines), we classify commits as being likely to have a fault, or unlikely to have a fault. Predictive accuracy figures for each project are typically in the mid-70's. Hence, it is possible for a configuration management system to inform a developer, post-commit, that they have just created a bug (with appx. 75% likelihood). We will also present an Eclipse plugin that can perform these predictions during a developer's editing session. Jim Whitehead is an Associate Professor of Computer Science at the University of California, Santa Cruz. Jim's research interests lies in the area of software evolution, software configuration management, application layer internet protocols, and software design. He has recently developed a new degree program in computer gaming, the BS Computer Science: Computer Game Design. Jim received his PhD in Information and Computer Science from UC Irvine, in 2000, under his advisor Richard N. Taylor. David Corina, UC Davis Evidence from the processing of American Sign Language. There is growing interest in the relationship between language and human action perception. The discovery of a mirror-neuron system (MNS) in Macaque monkeys provides a neural basis for a tight coupling between action and perception representations and has served as a basis for fertile speculation of an intimate relationship between human action perception and language. In this talk, I take a critical look at the claims of a mirror-neuron system in humans and discuss data from the processing of a manual language, American Sign Language and non-linguistic gestures. Our data calls into question the explanatory power of a MNS account of human language. David Corina holds a BS degree in Educational Psychology from New York University, an MA in linguistics from Gallaudet University and a Ph.D. in Psychology and Cognitive Science from UCSD in 1991. His work concerns the neural processing of language. He draws on comparisons from signed and spoken language processing as a means to elucidate core neural systems involved in human linguistic communication. Dr. Corina uses a wide arsenal of techniques in his studies, including, behavioral, neuropsychological, functional imaging, cortical stimulation mapping and electrophysiology. He is currently a professor in Depts. of Linguistics and Psychology and Faculty at the Center for Mind and Brain at U.C. Davis. He serves as the Scientific Co-Director of the NSF sponsored Science of Learning Center at Gallaudet University on Visual Language and Visual Learning (VL2). Philip Wolff, Emory University According to process theories of causation, people represent causation by modeling the physical and social processes that bring about causation in the world. These theories usually require that causal relations involve an uninterrupted chain of influences from the cause to the effect. A key problem for this view is the phenomenon of "negative causation." Negative causation is present when causation occurs in the absence of a cause. We say, for example, "The absence of nicotine causes withdrawal" or "Lack of water causes thirst." It is also present in cases of so-called "double prevention," or, situations where preventions are prevented, as when, for example, rescuers prevent guards from preventing an escape and thereby cause or allow the escape. In all cases of negative causation, there is a gap in the chain of influences from the cause to the effect. In my talk I show that negative causation is not, in fact, a problem for process theories based on force dynamics. Indeed, several patterns in the meaning of causal expressions encoding negativecausation may provide support for process approaches over competing approaches. According to statistical, counterfactual, and logical approaches to causation, expressions of causation involving negation and positive causation are symmetric: for example, NOT-CAUSE--> PREVENT and PREVENT -->NOT-CAUSE. In contrast, from a force dynamic perspective these different expressions are often related to each other asymmetrically: for example, NOT-CAUSE --> PREVENT, but not PREVENT --> NOT-CAUSE. The predictions of the force dynamic approach were supported in several experiments in which people re-expressed causal expressions taken from the internet and described animations depicting complex causal interactions. Because these asymmetries cannot be explained by statistical or logical approaches, the results support the view that causal reasoning involves simulating the actual processes that bring about causation in the world. Phillip Wolff received his Ph.D. at Northwestern University and is currently an assistant professor of Psychology at Emory University. His research focuses on the relationship between language and cognition, computational models of causal meaning and reasoning, and cross-linguistics approaches to the study of word meaning. His research includes the use of near-photorealistic animation to examine theories of causal meaning. Dr. Wolff has co-authored and edited two books and is currently co-editing a book with Barbara Malt entitled Words and the World, which examines the interface of language and thought across languages. Professor Wolff is on the editorial board of the journal Cognitive Science, and served as faculty at the 2007 Summer Institute of Linguistics at Stanford University. Roland Winston, UC Merced Some 50 years ago a group of brilliant astrophysicists figured out the nucleosynthesys of the elements in the interior of stars. Now, half a century later, with the benefit of advances in the structure of matter, we can ponder the consequences. Roland Winston came to UC Merced as a founding professor in the schools of Engineering and Natural Science from The University of Chicago, where he chaired the Physics Department. At UC Merced he has built a group in Solar Energy with a 4-acre testing laboratory at the former Castle Air Force Base. Ralph D. Freeman, UC Berkeley The presentation will focus on neurometabolic coupling in the cerebral cortex and the effect of electrical stimulation (TMS) on the process. Ralph Freeman is Professor of Vision Science and Optometry at UC Berkeley. He has an appointment in the Helen Wills Neuroscience Institute, and is affiliated with the Departments of Biophysics and Bioengineering at Berkeley. He is the recipient of many awards, including a NIH Research Career Development Award, and is Elected Fellow of the American Association for the Advancement of Science. Freeman has published extensively on the neural basis of binocular vision, biophysics and biochemistry of corneal tissue, and neurophysiological issues of development and plasticity in visual perception. His recent work focuses on connections between metabolic and neural factors in the cerebral cortex. Victor Zordan, UC Riverside Automatically creating human like animation for characters is difficult, especially in applications such as video games and online environments where the characters must move realistically, respond to unpredicted events, and remain controllable at a high level by the users of such virtual worlds. In this talk, I describe several techniques for generating realistic character motion using examples recorded from humans and physically based models, focusing primarily on controllable, responsive characters that combine dynamic simulation and recorded data. My research relies on human examples to dictate movement style and on simulation to create physically plausible motion including interactions with the environment and other simulated characters. Emphasis will be placed on generating believable unpredicted responses within a motion capture dependent animation system as well as on using both motion capture and simulation alone as mechanisms for generating high fidelity movement for humans. The talk will close with a brief discussion about the role of physics in generating games and online motion that is beyond the scope of applications seen today. Assistant Professor of Computer Science and Engineering at UC Riverside, Dr. Victor Zordan received his Ph.D. in computer science from Georgia Institute of Technology in 2002. Professor Zordan's research interests fall in several areas of computer animation including human motion, physically based modeling, interactive virtual environments, behavior control, and interface design. He has published numerous papers on the control of human and humanlike characters as well as on several other topics including anatomical modeling, procedural approaches, and video-based animation. Rich Ivry , UC Berkeley I will address the ancient problem of why is it so difficult to rub your stomach while patting your head. While this task is somewhat difficult to control in the laboratory, variants have been well-studied in the motor control literature, designed to provide insight into our competence and limitations in the production of actions. The emphasis in this literature has been on constraints related to motor programming and execution. I will offer an alternative framework, arguing that many of the constraints arise at a more cognitive level, reflecting the manner in which the task goals are represented. This perspective allows us to see how qualitative changes can emerge, both in terms of behavior and associated neural systems, between tasks involving subtle differences in the movements themselves. Professor Richard Ivry is Director of the Institute of Cognitive and Brain Sciences, and Director of the Cognition and Action Lab at UC Berkeley. He is faculty n the Department of Psychology and Helen Wills Neuroscience Institute. Ivry is the author of numerous articles and two books in cognitive neuroscience and cognitive psychology, including The Two Sides of Perception. He has served on several editorial boards, including Journal of Experiment Psychology: Human Perception & Performance, and Journal of Cognitive Neuroscience. Ivry’s research includes groundbreaking work on bimanual coordination, bimanual interference, hemispheric asymmetries in conceptual knowledge, and reaching movements. Stefano Carpin, UC Merced Started in 1997, Robocup has grown into one of the major yearly robotics events in the world. The spectrum of applications has also significantly expanded. Nowadays Robocup participants develop humanoid robots playing soccer games, as well as robotic platforms to assist first responders in the aftermath of major disasters. In this talk I will illustrate the major achievements obtained by this research community, and will detail some of the results I have produced while participating in the Robocup Rescue competitions since 2003. Stefano Carpin obtained the “Laurea” degree in Electrical Engineering and Computer Science in 1999, and the PhD in Industrial Electrical Engineering and Computer Science in 2003, both from the University of Padova, Italy. From 2003 to 2006 he held faculty positions at the International University Bremen, Germany, as research instructor and assistant professor of computer science. He joined UC Merced in January 2007 as assistant professor in the School of Engineering. His research interests are in the field of autonomous robotics, with a special emphasis on cooperative multi-robot systems and robot algorithms. He is an elected executive committee member of the Robocup Federation for the term 2007-2009. Songhwai Oh, UC Merced The data association problem arises in many areas of engineering Songhwai Oh is an assistant professor of Computer Science in the Seán Ó Nualláin, UC Berkeley While the focus of this talk is experimental epistemology, it inevitably encroaches on the issue of the disciplinary nature of cognitive science, and indeed the thorny issue of the academy and the world. After roughly half a century of cognitive science, asymptotic limits on the formal treatment of natural language are discernible; neuroscience tends to get stuck in local minima like FMRI; the issue of treating culture formally begs the question of how a discipline rooted in methodological solipsism can possibly countenance culture; in AI applications, we have yet to produce systems that are both embodied and symbolic. Meanwhile, in consciousness studies, as indeed elsewhere, the philosophical debates less frame the science than smother it. Though all this may already be apparent to veteran cognitive science practitioners, the direction taken in this talk may come as something of a surprise. It ends with a proposal for a coherent discipline that respects both the individual cognizer, animate or machine, and the real world. In summary, my goal is to demonstrate that by combining behavioral, neurophysiological, and computational modeling techniques we can gain substantial insights into the neural mechanisms underlying cognitive functions. Seán Ó Nualláin holds an M.Sc. in Psychology from University College, Dublin (UCD) Ireland & a Ph.D. in Computer Science from Trinity College, Dublin, Ireland. He is a visiting scholar at Walter Freeman's lab at UC Berkeley, coming there from a faculty position on the symbolic systems program at Stanford, and is due to join the Berkeley faculty in fall, 2007. He is the author of a book on the foundations of Cognitive Science: "The Search for Mind" (Ablex, 1995; 2nd ed Intellect, 2002; Third edition Intellect, 2003) and co-editor of "Two Sciences of Mind" (with Paul Mc Kevitt and Eoghan Mac Aogain) (Benjamins, 1997); editor of "Spatial Cognition"; co-editor of "Language, Vision, and Music" (Benjamins, 2002). His "Being Human: the Search for Order" (Intellect, 2002), which deals with science and society, sold out its first print-run immediately and has been published in a second edition (2004). He has worked the French jazz circuit and the American folk circuit between 2002 and 2007 as a guitarist with his partner Melanie O' Reilly after winning a landmark judgement preserving academic tenure in Ireland. He is a co-founder of the musician's union of Ireland, which now includes many of Ireland's leading musicians. Danny Oppenheimer, Princeton Oftentimes, people do not hold stable and consistent preferences and beliefs about the world. Rather, preferences and beliefs are typically constructed on the fly to conform to contextual cues. Because of this, people are surprisingly susceptible to manipulations which bias their preferences and judgments. This talk will explore the boundary conditions of these effects, and demonstrate how small and seemingly trivial changes in context can yield counterintuitive effects on reasoning. Danny Oppenheimer did his undergraduate work at Rice University and his Ph.D. at Stanford, both in psychology. He is currently on the faculty of Princeton University, where he holds a joint appointment in the Department of Psychology and the Woodrow Wilson School of Public Affairs. Dr. Oppenheimer has won several awards for his research, including most recently the 2006 IgNobel Prize in literature for his paper "Consequences of Erudite Vernacular Utilized Irrespective of Necessity: Problems with using long words needlessly". He is currently on sabbatical at UCLA. Jochen Ditterich, UC Davis Computational models based on diffusion processes have been proposed to account for human decision making behavior in a variety of tasks. In these human studies, however, the brain had to be treated as a “black box”. Here I explore whether such models account for the speed and accuracy of perceptual decisions in a reaction-time random dot motion direction-discrimination task and whether they can explain the decision-related activity of neurons recorded from the parietal cortex (area LIP) of monkeys performing the task. While a relatively simple diffusion model can explain the psychometric function and the mean response times, it fails to account for the response time distributions. By adding an “urgency mechanism” to the diffusion model, the psychometric function, the mean response times, and the shape of the response time distributions can be explained. Such an urgency mechanism could be implemented in different ways, but the best match between the physiological data and model predictions is provided by a diffusion process with a time-variant gain of the sensory signals. It can be shown that such a time-variant decision process allows the monkey to perform optimally (in the sense of maximizing reward rate). In summary, my goal is to demonstrate that by combining behavioral, neurophysiological, and computational modeling techniques we can gain substantial insights into the neural mechanisms underlying cognitive functions Jochen Ditterich is an assistant professor at the Center for Neuroscience and the Section of Neurobiology, Physiology & Behavior in the College for Biological Sciences at the University of California at Davis. He received his masters degree in electrical engineering and information technology (EEIT) in 1995 from the Technical University of Munich, Germany with a main focus on information processing in biological systems. He received his Ph.D. also in EEIT from the Technical University of Munich, Germany in 2000 with a dissertation titled “Saccade adaptation: dynamics, visual information processing, attention focus, and efference copy.” He has been at UC Davis since 2004 after finishing a postdoc at the Dept. of Physiology & Biophysics at University of Washington, Seattle. Dan Montello
, UC Santa Barbara The study of environmental cognition is the study of cognitive structures and processes that refer to the earth-surface surrounds in which we live and carry out our daily activities. Although the environment is recognized to include physical, social, and cultural aspects, research in this area usually focuses on the cognition of the physical environment, both natural and built (human-made). Environmental cognition includes both spatial and nonspatial properties, the cognition of which supports adaptive behavior, such as being oriented, moving efficiently, making rewarding choices while avoiding harm, and communicating about space and place. After I introduce and overview the multidisciplinary field of environmental cognition, I will exemplify some of its breadth and depth by briefly presenting three research projects conducted by my colleagues and me. The first concerns path angularity and spatial orientation, the second concerns individual differences in learning new places, and the third concerns vague cognitive regions (such as “downtown”). Dan Montello is a Professor in the Department of Geography at the University of California, Santa Barbara (UCSB). He is also Affiliated Professor in the Department of Psychology and in the Cognitive Science Graduate Emphasis Program, both at UCSB. Dan’s research is in the areas of spatial, environmental, and geographic perception, cognition, affect, and behavior; cognitive issues in cartography and GIS; and environmental psychology and behavioral geography. He earned his doctorate in Psychology (Environmental Psychology area) at Arizona State University and was a postdoctoral fellow at the Institute of Child Development at the University of Minnesota. Martin Banks, UC Berkeley When a picture is viewed from positions other than its center of projection, there can be large changes specified in the retinal image, yet the perceived spatial layout and shape of objects do not seem to change. We have shown that compensation for oblique viewing occurs provided that the viewer can estimate the slant of the picture surface accurately. Compensation is nearly veridical with binocular viewing at close range. Compensation generally does not occur with monocular viewing through a small aperture; instead, the percept is dictated by the shape of the retinal image. Our findings help to explain invariance for incorrect viewing positions, and other phenomena like perceived distortions with wide fields of view and the anamorphic effect. Our findings also have relevance to the design of displays. We will discuss, for example, how the viewer's position ought to affect percepts depending on the shape of the display surface. We are currently investigating similar phenomena with stereographic pictures and finding that quite different rules apply. Dr. Martin Banks is a Professor of Optometry and Vision Science and an Affiliate Professor of Psychology and Bioengineering at the University of California at Berkeley. He received a B.A. in Psychology from Occidental College, an M.A. in Experimental Psychology from the University of California at San Diego, and a Ph.D. in Child Psychology from the University of Minnesota. He was chairman of the Vision Science Program at UC Berkeley from 1995 through 2002. His research interests are in visual space perception and sensory combination. Charless Fowlkes, UC Berkeley In the 1920s, the Gestalt school identified grouping and figure-ground as two major principles underlying the process of perceptual organization. By the use of cleverly constructed stimuli, they were able to demonstrate the role of factors such as proximity, similarity, and curvilinear continuity in visual grouping and convexity, size, and symmetry in figure-ground assignment. Although these ideas were introduced more than 80 years ago, there have been several barriers to the centrality of perceptual organization in a computational theory of vision. It is seldom obvious how these cues apply to the diversity present in natural scenes, how conflicting cues are fused into a cohesive percept or even why a particular cue is used by the visual system in the first place. I will describe our work on the "ecological statistics" of perceptual organization which provides a fresh outlook on these difficulties by connecting the existence of such cues to the statistics of the natural world. Charless Fowlkes received a B.S. in Engineering and Applied Sciences from Caltech in 2000 and a Ph.D. in Computer Science from U.C. Berkeley in 2005. He is currently a post doctoral research at the Lawrence Berkeley National Laboratory. He works on statistical approaches to perceptual organization in low and mid-level vision as well as developing computational tools for understanding morphology and spatial patterns of gene expression in animal development. Shumin Zhai , IBM Almaden Research Center In this presentation I will give a survey of three areas of research at the IBM Almaden Research Center, all concerning input actions on the computer screen: 1) Laws of action -- pointing, crossing, and steering: To expand the scientific foundation of user interface research, it is necessary to uncover human action regularities in interacting with computers. I will comment on using Fitts' law to characterize pointing performance, followed by reviewing performance regularities in two different categories of movement relevant to user interface design and evaluation - goal crossing and path steering (with Johnny Accot). 2) Contextual eye-tracking based interaction: Looking into the future, computer vision and other sensing technologies will play an important role in human-computer interaction. I will discuss how eye-gaze is used as an implicit contextual source of information in MAGIC pointing (with Carlos Morimoto and Steve Ihde), and similarly how eye-gaze information can augment multi-modal human-machine dialogue systems in iTourist (with Pernilla. Qvarfordt). 3) High performance mobile input: Mobile phones are expected to play increasingly central role in personal computing. Developing effective user interfaces in mobile forms presents an opportunity and a challenge for the HCI field. I will describe and demonstrate ShapeWriter --- using geometric shapes defined on a keyboard layout to efficiently enter text (with Per-Ola Kristensson). Shumin Zhai is a Research Staff Member at the IBM Almaden Research Center. Named a Distinguished Scientist by ACM in 2006, his work has produced about 100 papers, received numerous patents, contributed to three IBM Research Division Accomplishments, and been broadly reported in the news media. He is on the editorial boards of Human-Computer Interaction, ACM Transactions on Computer-Human Interaction, and other journals. He has been a visiting professor and lectured at various universities in the US, Europe and China. He earned his Ph.D. degree at the University of Toronto. Jonathan Winamer, MIT What happens in our brains during imagination? The degree to which mental imagery recruits the same neural structures and processes as viewing real visual scenes has been a matter of much debate. Neuroimaging research has shown that early perceptual brain areas can be recruited during imagery, but do imagery and perception use the same neural circuits to represent the same visual features? I will discuss a series of experiments in which we made use of the motion aftereffect (MAE) to address this question. The experiments show that imagining moving patterns, as well as viewing static photographs with implied motion, yielded direction-specific MAEs on real motion test probes. The transfer of adaptation from imagined and implied motion to perception of real motion demonstrates that at least some of the same direction-selective neurons involved in imagining motion and inferring motion are also used for perceiving actual motion. Jonathan Winawer is currently a PhD student in Brain and Cognitive Sciences at the Massachusetts Institute of Technology. He received a BA in Classics from Columbia University and a BS in Neurobiology from the City College of New York. His advisors at MIT are Bart Anderson and Lera Boroditsky, and his Phd studies have focused on visual perception and visual cognition, particularly lightness perception, synesthesia, language and color discrimination, mental imagery, and face representation and adaptation. He expects to graduate this summer. Michael Spivey, Cornell (and soon, UC Merced) Rather than a sequence of logical operations performed on discrete symbols, real-time cognition is better described as continuously changing patterns of neuronal activity. The continuity in these dynamics indicates that, in between describable states of mind, much of our mental activity does not lend itself to the linguistic labels relied on by much of psychology. I will discuss eye-tracking and computer-mouse-tracking evidence for this temporal continuity in speech perception, sentence processing, categorization, and decision-making. I will also provide geometric visualizations of mental activity depicted as a continuous trajectory through a neuronal state space. In this theoretical framework, close visitations of labeled attractors may constitute word recognition events and object recognition events, but the majority of the mental trajectory traverses unlabeled regions of state space, resulting in multifarious mixtures of mental states. Michael Spivey is Director of the Cognitive Science Program and Associate Professor of Psychology at Cornell University. He received his PhD in Brain and Cognitive Sciences from the University of Rochester in 1996, and has been a professor at Cornell since then. In 2008, he will move to UC Merced as Professor of Cognitive Science. His research explores the continuous transitions that the mind traverses as it changes from one state to the next (whether those states are phonemes, words, visual objects, or categories). In addition to his recent book, The Continuity of Mind (Oxford U. Press), Spivey has published over 35 journal articles, and over 50 book chapters and conference proceedings papers demonstrating the fluid interaction over time exhibited by syntax, semantics, and visuospatial information. Paul Skokowski, Stanford Frank Jackson's Knowledge Argument is a powerful argument against some forms of materialism, in particular mind-brain identity theories. But there are other versions of materialism that Jackson's argument misses entirely. I'll explain why in this talk. Along the way, I’ll give a brief history of the mind. Biography: Paul Skokowski currently teaches in Symbolic Systems at Stanford and Philosophy at UC Berkeley. He has been a McDonnell-Pew Fellow at the Centre for Cognitive Neuroscience at Oxford University, and was previously Director of the Institute for Scientific Computing Research at the University of California Lawrence Livermore National Laboratory, and Manager of Search worldwide at Yahoo! Inc. He has a B.A. in Physics and Philosophy from Oxford University and a Ph.D. in Philosophy from Stanford University. John Funge, AI Live In this talk I will give a quick introduction to artificial intelligence in games (Game AI). Followed by a short presentation about the prefered method of animation in games and the sometimes surprising interdepenencies between animation and AI. I will end by briefly mentioning some ground breaking products that I helped develop Biography:John Funge (www.jfunge.com) is a co-founder and one of the lead scientists at AiLive Inc. At AiLive he is part of the team that developed LiveMove and LiveCombat, two commercial products that bring ground breaking real-time machine learning technology to the computer entertainment industry. John previously worked at Sony Computer Entertainment America's (SCEA) research lab. He received his Ph.D. in Computer Science from the University of Toronto and also holds degrees from the University of Oxford and King's College London. John is the author of numerous technical papers and two books on Game AI, including his latest book "Artificial Intelligence for Computer Games: Michael Spivey, Cornell Rather than a sequence of logical operations performed on discrete symbols, real-time cognition is better described as continuously changing patterns of neuronal activity. The continuity in these dynamics indicates that, in between describable states of mind, much of our mental activity does not lend itself to the linguistic labels relied on by much of psychology. I will discuss eye-tracking and computer-mouse-tracking evidence for this temporal continuity in speech perception, sentence processing, categorization, and decision-making. I will also provide geometric visualizations of mental activity depicted as a continuous trajectory through a neuronal state space. In this theoretical framework, close visitations of labeled attractors may constitute word recognition events and object recognition events, but the majority of the mental trajectory traverses unlabeled regions of state space, resulting in multifarious mixtures of mental states. Michael Spivey is Director of the Cognitive Science Program and Associate Professor of Psychology at Cornell University. He received his PhD in Brain and Cognitive Sciences from the University of Rochester in 1996, and has been a professor at Cornell since then. In 2008, he will move to UC Merced as Professor of Cognitive Science. His research explores the continuous transitions that the mind traverses as it changes from one state to the next (whether those states are phonemes, words, visual objects, or categories). In addition to his recent book, The Continuity of Mind (Oxford U. Press), Spivey has published over 35 journal articles, and over 50 book chapters and conference proceedings papers demonstrating the fluid interaction over time exhibited by syntax, semantics, and visuospatial information. Arthur Markman , UT Austin There is a tendency within Cognitive Science to seek a single representation framework that accounts for most (if not all) cognitive processing. I have argued that there are many different systems of representation, and that they each support different kinds of processes. Thus, the representation system used by a particular aspect of cognitive processing must be matched to the information that must be available for a particular cognitive task. I illustrate this principle by focusing on the role of discrete and continuous representations in cognitive systems. I argue that discrete representations are critical for helping systems communicate and establish reference. Arthur B. Markman is Annabel Irion Worsham Centennial Professor of Psychology and Marketing at the University of Texas at Austin. He received his PhD from the University of Illinois in 1992, and has also held positions at Northwestern University and Columbia University. His research examines similarity and analogy, categorization, decision making, motivation, and knowledge representation. He has published over 100 scholarly works including the book Knowledge Representation, as well as a Cognitive Psychology textbook coauthored with Douglas Medin and Brian Ross. Dr. Markman served as Executive Officer of the Cognitive Science Society from 2001-2003, and is currently editor of the journal Cognitive Science. Marcelo Kallmann , UC Merced In this talk I will present an overview of the algorithms and Marcelo Kallmann is Assistant Professor and Founding Faculty at the David C. Noelle, UC Merced Formal cognitive neuroscience models of how we learn associations and David C. Noelle has recently joined the faculty of the University of Nick Davidenko, Stanford Face perception is an essential and remarkable human ability that has yet to be characterized at a representational level. A major avenue of progress for face perception research has been the study of disparities in our capacity to represent different faces in memory. This talk will Nicolas Davidenko recently finished his Ph.D in the Psychology Teenie Matlock, UC Merced Fictive motion sentences such as "The road runs along the coast" are ubiquitous in English and many other languages. They are especially common when people are describing spatial layouts. These constructions have been of interest to linguists for some time because they include a motion verb but describe no motion. On one view, the conceptual structure of these sentences is static, and not unlike that of non-fictive motion sentences, such as "The road is next to the coast" (Jackendoff, 2002). On another view, it is dynamic, including tacit mentally simulated motion (Talmy, 1996). Using a variety of experimental methods, such as reading comprehension, eye-tracking, drawing, and natural discourse, I have investigated the understanding and use of these sentences. In my presentation, I will discuss the main results of my work and their implications for mental representation. Teenie Matlock is Assistant Professor of Cognitive Science and founding faculty at UC Merced. She has published articles on language and cognition in cognitive science, linguistics, and psychology. Her primary research interests are spatial language, abstract concepts, metaphor, gesture, and conceptual semantics. She did graduate studies in linguistics at UC San Diego, and finished her PhD in cognitive psychology at UC Santa Cruz. She was a post-doctoral researcher in the Psychology Department at Stanford University. Daniel Richardson, UCSC I will describe a series of eye tracking projects that seek to understand how people direct and coordinate their attention during naturalistic language use. We present pictures, figurative speech, and videos to participants, who watch the displays, form opinions, have discussions and play games. In the first project, we examined how people process figurative speech and other forms of implicit spatial language. In the second we investigated how people look at members of a minority group when potentially offensive remarks are spoken. In the third project we quantified the temporal coupling between two people’s eye movements. We eye tracked conversants simultaneously while they talked about TV, art, politics and matched ambiguous figures. Our results demonstrate that visual and verbal processes are tightly linked, and reveal the role of common ground and the coordination of visual attention during joint activities. Daniel studied philosophy at Magdalen college, Oxford University, and then psychology at Cornell. After receiving his PhD, he became a postdoctoral researcher and lecturer at Stanford. Since this year, he has been an assistant professor at UC Santa Cruz. Jerome Feldman, UC Berkeley The neural revolution in cognitive science, which was always inevitable, is well under way. Rapid advances on the neural correlates of intelligent behavior have transformed cognitive experiment and theory. There is already enough known about how our brains process information to render many traditional theories obsolete and a unified neurally-based cognitive science is arising. Linguistics and Philosophy have, for both historical and technical reasons, been slow to integrate even the most basic neuroscience. Much of fundamental neuroscience is done with animals and, since only people use language, there is no easy way to extend animal findings to human thought and language. This new book is the first systematic attempt to show how human language and thought arise as an extension of the physiology and experiences that people share with other animals. Integrating findings from all the cognitive sciences yields a foundation for an explicitly neural theory of language that is an integral part of contemporary science. Many (but not all) of the deep questions concerning mind and brain are less mysterious when expressed as neural computation. Jerome A. Feldman is Professor of Electrical Engineering and Computer Science and of Cognitive Science at the U. California at Berkeley and a Bruno Olshausen, UC Berkeley Our percepts of the world are clearly inferred, rather than being computed directly from the available data. This means that our brains must be endowed with powerful inferential machinery - i.e., probabilistic models - for combining incoming sensory information together with prior knowledge in order to infer what's "out there" in the environment. In this talk I will present a simple version of a probabilistic model for primary visual cortex (V1) that is based on the idea of sparse coding - i.e., where images are represented by a small number of active units at any given time. I will then present the results of computational simulations showing that this idea is consistent with the receptive field properties found in V1 neurons, and I will present data supporting the idea that cortical neurons are attempting to infer sparse representations of images. Both the model and the data make clear that if we are to actually understand what is going on the cortex, we need to focus our efforts on studying how it operates under natural conditions. Dr. Bruno Olshausen is currently an associate professor in the Helen Wills Neuroscience Institute and School of Optometry at UC Berkeley. He earned BS and MS degrees in Electrical Engineering from Stanford University, and a PhD degree in Computation and Neural Systems from the California Institute of Technology. Dr. Olshausen has been a postdoctoral fellow in the Department of Psychology at Cornell University; a postdoctoral fellow at the Center for Biological and Computational Learning at the Massachusetts Institute of Technology; an assistant and then associate professor in the Department of Psychology and Center for Neuroscience at UC Davis; an associate professor in the Department of Neurobiology, Physiology & Behavior, and Center for Neuroscience at UC Davis; and a principal investigator at the Redwood Neuroscience Institute in Menlo Park, California. Petros Faloutsos, UCLA In this talk, I will present an overview of the research that the MAGIX group is focusing on at UCLA. Our research covers many aspects of human character animation such as: Petros Faloutsos is an assistant professor at the Department of Computer Science at the University of California at Los Angeles. Faloutsos received his PhD degree (2002) and his MSc degree in Computer Science from the University of Toronto, Canada and his BEng degree in Electrical Engineering from the National Technical University of Athens, Greece. Professor Faloutsos research interests include most aspects of human character simulation, and rendering. He is the founder and the director of the graphics lab at the Department of Computer Science at UCLA. He has served as the rogram co-chair of the 2005 ACM SIGGRAPH/Eurographics Symposium on Computer Animation. Professor Faloutsos is among the 10000 most cited authors in Computer Science according to Citeseer 2005. He has received a 2002 Okawa Foundation Research Grant and a 2001 best paper award from Computers & Graphics. Professor Faloutsos is a member of the ACM and the Technical Chamber of Greece (Professional Engineers). Dominic Massaro, UCSC Speech and language science and technology evolved under the assumption that speech was a solely auditory event. However, a burgeoning record of research findings reveals that our perception and understanding are influenced by a speaker's face and accompanying gestures, as well as the actual sound of the speech. Perceivers expertly use these multiple sources of information to identify and interpret the language input. Given the value of face-to-face interaction, our persistent goal has been to develop, evaluate, and apply animated agents to produce realistic and accurate speech. Baldi is an accurate three-dimensional animated talking head appropriately aligned with either synthesized or natural speech. Baldi has a realistic tongue and palate, which can be displayed by making his skin transparent. Based on this research and technology, we have implemented computer-assisted speech and language tutors for children with language challenges and persons learning a second language. Our language-training program utilizes Baldi as the conversational agent, who guides students through a variety of exercises designed to teach vocabulary and grammar, to improve speech articulation, and to develop linguistic and phonological awareness. Some of the advantages of the Baldi pedagogy and technology include the popularity and effectiveness of computers and embodied conversational agents, the perpetual availability of the program, and individualized instruction. The science and technology of Baldi holds great promise in language learning, dialog, human-machine interaction, education, and edutainment. Dominic W. Massaro is Professor of Psychology and Computer Engineering, director of the Perceptual Science Laboratory, and Chair of Digital Arts and New Media M.F. A. program at the University of California, Santa Cruz. He received a BA in Psychology (1965) from UCLA and an MA (1966) and a Ph.D. (1968) in Psychology from the University of Massachusetts-Amherst. After a two-year postdoctoral fellow at the University of California, San Diego, he was a professor at the University of Wisconsin until 1979 before moving to Santa Cruz. He has been a Guggenheim Fellow, a University of Wisconsin Romnes Fellow, a James McKeen Cattell Fellow, and an NIMH Fellow. He is a past president of the Society for Computers in Psychology, and is currently the book review editor of the American Journal of Psychology and founding co-editor of the journal Interpreting. He has published numerous academic journal articles, written and edited several books (including Perceiving talking faces: from speech perception to a behavioral principle, Cambridge, Massachusetts: MIT Press; The Science of the Mind: 2001 and Beyond, New York: Oxford University Press; and Experimental Psychology: An information processing approach, Orlando, FL : Harcourt Brace Jovanovich.). His research uses a formal experimental and theoretical approach to the study of speech perception, reading, psycholinguistics, memory, cognition, learning, and decision-making. One focus of his current research is on the development and theoretical and applied use of a completely synthetic and animated head for speech synthesis, language tutoring, and edutainment. Rafael Núñez , UCSD How can we "objectively" share abstract entities with others, in a stable Rafael Núñez is an Associate Professor at the Department of Cognitive Rick Grush, UCSD A typical view of the temporal content of perception is that it mirrors the temporal features of what is being perceived. There are two aspects to this view. First, that at any instant the temporal content of experience is a corresponding instant. And second, that the representation of succession and simultaneity is accomplished through the succession and simultaneity of representations. I will overview philosophical considerations that cast doubt on the first, and psychological considerations that cast doubt on the second. I will then briefly outline a novel account of the relevant information-processing structure of perception that accounts for the phenomena, an account that explains how it is that at any instant the content of perceptual content is a temporal interval, and the specifics of this interval are in part a matter of active interpretation, not merely passive reflection, of the temporal features of the perceived environment Rick Grush completed a joint doctorate at UCSD in 1995 in Philosophy and Cognitive Science, after which he held post-doctoral positions at the Philosophy-Neuroscience-Psychology program at Washington University in St Louis, and the Center for Semiotic Research at the University of Aarhus, Denmark. From 1998-2000 he was in the Philosophy Department at the University of Pittsburgh, where he had secondary appointments in Linguistics and the Intelligent Systems Program, and was a member of the Center for the Neural Basis of Cognition. Amy Pritchett, GIT Abstract Amy Pritchett is the David D. Lewis Associate Professor of Cognitive Engineering in the School of Aerospace Engineering and a joint Associate Professor in the School of Industrial and Systems Engineering at the Georgia Institute of Technology. Her research encompasses human-automation interaction, including advanced decision aids; procedure design as a mechanism to define and test the operation of complex, multi-agent systems (e.g. air traffic control, spacecraft mission control); simulation of complex systems to assess changes in emergent system behavior in response to implementation of new information technology; and educational technology and the design of educational systems. She is on the editorial board of the Journal of Cognitive Engineering and Decision Making, an area editor of the Transactions of the Society for Computer Simulations, and associate editor of the AIAA Journal of Aerospace Computing, Information and Communication. Her awards include the RTCA Jackson Award for contribution to aviation. Dr. Pritchett is currently a member of the National Academies' Aeronautics and Space Engineering Board. |