|
Center for Cognitive Science The Puzzle of the Mind |
|||||||||||||||||||||||||||||||
| Spring
1998 Colloquium
|
|||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||
| Regular
colloquia are Wednesdays, 2:00-4:00 p.m., 280 Park Hall, North (Amherst)
Campus, and are open to the public. Refreshments are served.
Other colloquia and our distinguished speaker are on different days.
ABSTRACTS
Wednesday,
January 21, 1998 JAMES F. ALLEN Department
of Computer Science "CONVERSATIONAL
PLANNING AGENTS: A conversational agent is a system that can engage in natural language conversation in order to further its goals. In our TRAINS project, the goals consist of simple planning and scheduling tasks, and the system supports mixed initiative planning using several modes of communication: spoken and keyboard natural language, mousable map displays, and menus. The project represents a concerted effort to bring work in natural language understanding, dialog modeling, and planning into a single coherent system. The TRAINS system is a specific agent that can support unconstrained dialogue in order to assist the user in solving randomly generated route planning tasks. Significant effort has been made to make the system robust, so that it can perform well even in the face of inevitable speech recognition errors, and can continue the dialogue in a natural way under any circumstances. This talk will discuss the overall project and its accomplishments so far, and then focus on a few specific mechanisms that enable robust dialogue behavior. Because the system can interact in close to real time, it provides a firm basis for experimentation and data collection to study different strategies for human-computer dialogue for collaborative problem solving.
Refreshments will be available
Wednesday, January
28, 1998 ALAN LOCKWOOD Depts.
of Neurology, Nuclear Medicine, Communication Disorders and Sciences,
and "FUNCTIONAL IMAGING OF THE HUMAN AUDITORY SYSTEM AND AUDITORY PHENOMENA" We have used positron emission tomography to map the human auditory stimuli. Stimulating the auditory system with unilateral tones at 500 and 4000 Hz elicits a systematic response in brainstem, midbrain, thalamic, and cortical regions. Our study provides evidence for the existence of tonotopic and ampilotopic organization of cortical portions of the central auditory system. Acitvity in the posterior and mid portions of the cingulate gyrus (BA 23) at low but not high intensity levels of stimuli may be evidence that this brain region functions as a gain control center. In studies of patients with tinnitus, we found evidence for spontaneous neural activity in the central auditory system that is not of cochlear origin. In addition, tinnitus patients have abnormal links between the auditory and limbic system that may account for the emotional impact of tinniuts. Finally, tinnitus patients with hearing loss express evidence for plastic transformations of the central auditory system. We hypothesize that tinnitus may be the auditory system analog of phantom limb sensations and that the nature and severity of tinnitus may be dependent on the extent of the plastic changes. Refreshments
will be available
Wednesday, February 4, 1998
2:00-3:30 p.m. 280 Park Hall North Campus PHIL MERIKLE Psychology, University of Waterloo "What
Experimental Studies of Perception Perception can occur without the subjective awareness of perceiving. In what ways are perception without awareness and perception with awareness similar or dissimilar? This question has been addressed in several ways. One approach has been to induce conscious people to perceive unconsciously by presenting stimuli under suboptimal conditions (e.g., short durations). A second approach has been to induce conscious people to perceive unconsciously by presenting stimuli outside their focus of attention. A third approach has been to study people in whom unconsciousness has been induced directly through the administration of surgical anesthesia. Regardless of how unconsciousness is induced, stimuli can produce qualitatively different consequences depending on whether they are perceived with or without awareness. The studies reveal that consciously perceived stimuli can guide intentional actions whereas unconsciously perceived stimuli lead to more habitual reactions which may even interfere with on-going intentional actions. These findings provide empirical support for the widely held assumption that being conscious allows us to act on the world and to produce effects on the world. All
are invited to attend
Wednesday,
February 18, 1998
2:00-3:30 p.m. 280 Park Hall North Campus DAVID NOELLE Center for the Neural Basis of Cognition "Modeling
An Interference Effect In Instructed Category learning is often seen as a process of inductive generalization from a set of class-labeled exemplars. Human learners, however, often receive direct instruction concerning the structure of a category before being presented with examples. Such explicit knowledge may often be smoothly integrated with knowledge garnered by exposure to instances, but not always. Some interference effects have been observed. Specifically, errors in instructed rule following may sometimes arise after the repeated presentation of correctly labeled exemplars. Despite perfect consistency between instance labels and the provided rule, such inductive training can drive categorization behavior away from rule following and towards a more prototype-based or instance-based pattern. In this talk, I will present a general connectionist model of instructed category learning which captures this kind of interference effect. I model direct instruction as a sequence of inputs to an artificial neural network which transforms such advice into a modulating force on classification behavior. Exemplar-based learning is modeled in the usual way for connectionist networks: as weight modification driven by error feedback. The proposed network architecture allows instruction following and exemplar-based learning to interact in a psychologically plausible manner. Simulation results are provided on a simple instructed category learning task, and these results are compared with human performance on the same task. All
are welcome to attend
Wednesday,
February 25, 1998
2:00-3:30 p.m. 280 Park Hall North Campus EVAN THOMPSON Department
of Philosophy "FINDING OUT ABOUT FILLING IN" Abstract-Visual scientists use the terms "filling in" and "perceptual completion" to refer to situations where subjects report that something is present in a particular region of visual space when it is actually absent from that region, but present in the surrounding area. Whether there is neural filling-in, however, is a matter of great debate. I will first present a taxonomy of perceptual completion phenomena to organize the discussion and then argue for the following points: (1) Certain forms of perceptual completion seem to involve spatially propogating neural activity (neural filling in) and so, contrary to Daniel C. Dennett's position, cannot be described as the results of the brain's "ignoring an absence" or "jumping to a conclusion." (2) Nevertheless, neural filling-in does not entail either "analytic isomorphism" or "Cartesian materialism." (3) Most discussions of filling-in have been based on a representational conception of vision. To reject this conception in favour of an "enactive" or "animate" conception reduces the importance of filling-in as a theoretical category in the explanation of vision. (4) The evaluation of perceptual content should not be dictated by "subpersonal" considerations about internal processing, but rather by considerations about the task of vision at the level of the animal or person interacting with the world. All
are invited to attend
Wednesday,
March 4, 1998
280 Park Hall 2:00 - 3:30 p.m. North Campus STUART C. SHAPIRO Department
of Computer Science A PROCEDURAL SOLUTION TO THE UNEXPECTED HANGING AND SORITES PARADOX
The paradox of
the Unexpected Hanging, related prediction paradoxes, and the sorites
paradoxes all involve reasoning about ordered collections of entities:
days ordered by date in the case of the Unexpected Hanging; men ordered
by the number of hairs on their heads in the case of the bald man
version of the sorites. The reasoning then assigns each entity a value
that depends on the previously assigned value of one of the neighboring
entities. The final result is paradoxical because it conflicts with
the obviously correct, commonsensical value. The paradox is due to
the serial procedure of assigning a value based on the newly assigned
value of the neighbor. An alternative procedure is to assign each
value based only on the original values of neighbors---a parallel
procedure. That procedure does not give paradoxical answers.
All are invited to attend
Wednesday,
March 18, 1998
280 Park Hall 2:00 - 3:30 p.m. North Campus Department
of Anthropology "Current
Anthropological Research After Berlin
and Kay (cir. 1975) found neurally determined universals in color
categorization, blatant cross-cultural differences remained to be
explained. These consisted of different stages of color-cateogry
evolution, systems of categorizing color that did not fit the neurally-based
model, plasticity and change among color categories, and category-internal
semantic relations previously overlooked. Such observations were
enhanced by expansion of worldwide and regional surveys and by an
improvement of Munsell-chip interviewing methods.
All are welcome to attend
Wednesday,
March 25, 1998
2:00 - 3:30 p.m. 280 Park Hall North Campus GAIL MAUNER Department
of Psychology and Center for Cognitive Science "HOW
DO WE KNOW THAT THE SHIP
Typically, our understanding of a sentence like "The ship was sunk" includes an unexpressed agent who is responsible for sinking the ship. Within the psychological literature, it is commonly assumed that this unexpressed information is derived from stored conceptual knowledge. An alternative view is that unexpressed agents are derived from linguistic sources such as lexical representations and implicit grammatical knowledge. In this talk, I will present recent data from experiments using eye-monitoring, cross-modal naming, and word-by-word reading to argue that (1) the unexpressed agent that is part of our understanding of a sentence like "The ship was sunk" is derived from linguistic sources; and (2) this information is accessed as soon as a reader recognizes the main verb in an agentless passive sentence. I will end by discussing the implications these data hold for theories of modularity in language processing.
Wednesday,
April 1, 1998
280 Park Hall North Campus 2:00 -3:30 p.m. BARRY SMITH
Department of Philosophy Introductory
Tutorial on On April 24-25 the Center will co-sponsor a major international conference in Buffalo on the topic of "Applied Ontology: Law and Institutions in Society". The first day of the conference will be devoted to John Searle's book The Construction of Social Reality, and John Searle will himself be present. As part of the preparations for this conference Barry Smith will give an introductory tutorial on Searle's cognitive theory of social ontology. All
are invited to attend
Wednesday,
April 1, 1998
280 Park Hall North Campus 2:00 -3:30 p.m. Corinne Jorgensen School of Information and Library Studies, SUNY Buffalo "Image Attributes: An Investigation" .
With the rapid expansion in imaging technologies, access to collections of digital images is a subject of major interest. Indexing systems and computerized retrieval for images both need data concerning typically described image attributes. To date, there is little research upon which to base choices as to which attributes should be included in these systems. This research is investigating attributes typically described in several types of tasks using pictorial images. Participants performed descriptive, categorizing, and searching tasks, and word and phrase data were subjected to content analysis. Forty-two image attributes and nine higher level attribute classes were described. The data suggest that indexing of literal object is of prime significance, as is indexing of the human form and other human characteristics. Content/Story and other abstract attributes are also typically described, suggesting that image indexing may benefit by application of concepts associated with indexing of fiction. Term variability is less than might have been expected, suggesting some constraints may exist on the process of communicating about visually perceived data. The data are of interest in terms of what they suggest about human perception of and communication about complex images. Time: Pizza and drinks at 5:30; informal talk 6-ish
Wednesday,
April 15, 1998
2:00 - 3:30 p.m. 280 Park Hall North Campus JANET DEAN FODOR CUNY Graduate Center "WHAT IS A PARAMETER?" Recent research in learnability has made it clear that for natural language there can be no instant "automatic" triggering of parameters. This is because the trigger properties in natural languages are often deep properties, not recognizable without parsing the input sentence. Current approaches such as Gibson and Wexler (LI 1994) therefore use the parsing routines to identify triggers. Unfortunately, the proposed mechanism for doing so if very inefficient. I show that this is because it does not respect the Parametric Principle: it evaluates millions of particular grammars, rather than establishing the values of 20 or 30 parameters. By tracking out why this is so, I have found a remedy for it. There is a way of using the parser that does implement the Parametric Principle. But it calls for a new conception of parameters and of their triggers: they are one and the same thing, and consist of features of small treelets, made available by UG and adoptable into individual grammars. This conclusion is in accord with most current theories of syntactic parameterization, including the Minimalist program, HPSG and TAG theory. All
are welcome
Wednesday,
April 22, 1998
2:00 - 3:30 pm 280 Park Hall North Campus ROBERTO CASATI NRS, Aix-en-Provence "WHAT SOUNDS ARE" The physicalist reduction of sounds seems to be achieved once we identify sounds with sound waves in a medium in which a sounding object (and possibly a hearer) is present. This raises some objections and leaves some questions open. Among other things, the relationship between the sound and the sounding object remains in the dark. I shall suggest that there is another and more promising candidate for the physical identification of sounds, namely processes or events inside (or at the surface of) sounding objects, or in the stuff of the sounding object. Three main issues will be discussed in the talk: the location of sounds, their nature, and the perceptual representation we have of them. All
are welcome to attend
Wednesday,
April 29, 1998
2:00 - 3:30 p.m. 280 Park Hall North Campus INGVAR JOHANSSON Department
of Philosophy "The Ontology of Patterns" Most ontologists regard patterns as some kind of easily analyzed subcategory, a mere by-product in need of no special attention. This view I will challenge. I think that, in fact, pattern belongs to the top-level categories, and that a close scrutiny of its specifica differentia may be of interest both for ontologists and for cognitive scientists. I will, as a starting point for discussion, list certain features which accrue to patterns: the feature of fiatness, the feature of non-reducibility to a set or aggregate of properties, parts or relations, the feature of having that kind of objective existence which can be represented cognitively (via language and perception). Reference will also be made to the question whether all patterns can be represented pictorially. All
are welcome |
|||||||||||||||||||||||||||||||
You are
visitor
to this
site.
Last updated on March 12, 2003 by H. Jones
Contact:
hhjones@buffalo.edu
The Center for Cognitive Science, University at Buffalo, State University
of New York, 652 Baldy Hall, Buffalo, NY 14260
Phone: (716) 645-3794, Fax: (716) 645-3825
© Copyright 2003, Center for Cognitive Science, University at Buffalo, All Rights Reserved.