6 October - Steve diPaola, School for Interactive Arts and Technology, Simon Fraser University

Computationally modeling the human creative and artistic processes

Using computer modeling techniques, scientists are gaining insight into how creativity works and how it is exploited by artists. Steve DiPaola will demonstrate his work recently published in the journals Nature and Leonardo showing how portrait masters intuit knowledge of how the science of vision and perception work and exploit this knowledge to guide the viewers eye and create a narrative in a portrait. Using eye tracking and computer modeling, the study specifically shows convincing evidence that Rembrandt used these techniques in his late portraits. He will also discuss his recent work showcased at MIT and Cambridge, using computer artificial intelligence to model human creative thought as a technique to better understand how creative minds think differently to solve problems.

For a profile of Steve DiPaola, and more about his work, please see:

Steve's School for Interactive Arts and Technology page
Steve's own website

20 October - Rick Grush, visiting professor in Department of Philosophy, Simon Fraser University (from University of California, San Diego)

Control, embodiment, and demonstrative semantics?

I present (work done in collaboration with Amanda Brovold) an account of the semantics of demonstratives, both exophoric and endophoric. The account, which represents a significant shift from accounts typical in philosophy and linguistics, has two features. First, it is argued that the conceptual archetype of what Langacker has called the control cycle is the key to understanding core demonstrative semantics. The control cycle is a conceptual archetype with many semantic and grammatical ramifications in which an agent recognizes something as a candidate for control, acts on it to bring it under control, and then controls it. This cycle can manifest in physical, perceptual, attentional, epistemic, and social domains. We analyze demonstrative semantics as keying on the interlocutors’ assessments of their own and each other’s (individual and joint) control domains – which objects they have control of, or are trying to gain, relinquish, or transfer control of. In languages with a proximal/distal distinction, proximals key on situations where the speaker S has control of the referent (and is often putting it in the addressee’s A control), and distals for referents S does not control.

This account i) does a better job explaining exophoric demonstrative choice than the spatial proximal/distal spatial interpretation; ii) provides a superior and unified interpretation of diverse experimental results; iii) provides a unified treatment of exophoric and endophoric anaphora; and iv) provides revealing analyses of how demonstratives pattern when different control cycles are involved.

(i)-(ii) Despite the view that ‘this’ is for things closer to S (speaker), and ‘that’ for things closer to A (audience), one can say ‘what’s this?” when touching a wound on someone’s back, even if one is touching it with a meter stick. Though closer to A spatially, S has control in several ways in which A does not. Also, as Coventry et al. (2008) have shown, giving S tools that extend the spatial range of their control also expands the range of things referred to with ‘this’ rather than ‘that’. These authors also found that demonstrative choice was affected by who last touched the referent – a subliminal indicator of social control (ownership).

(iii) Just as ‘this’ is preferred for things S has physical control of, and is placing in A’s control, so ‘this’ is preferred for ideas or propositions which S has attentional and epistemic control over, and is giving A control over (by telling her), e.g. 1a/1b.

(iv) Patterns can change as the relevant control domain changes. E.g. the pattern in 2a switches if A has physical control of S’s wrench, and S is giving A social control (ownership) of the wrench that A already has physical control of. An analogous switch occurs in 2c. In 2b/d, the pattern assumes that only the potential joke teller knows (and hence has epistemic control over) the joke. However, if S is given (metonymically) physical/perceptual control, e.g. physically touching one entry one a written list of jokes, ‘this’ is licensed. And if S does not know the joke, but points at a list, ‘that’ is licensed in 2b.

We explain what to make of languages whose demonstratives have specific spatial import (e.g. Burnhult 2008). And we respond to the potential charge that the notion of control is too vague, or is subsumed under notions such as accessibility or a metaphorical understanding of proximity.

The second feature is the embodied nature of exophoric demonstrative reference. As argued in the first part, demonstrative semantics keys on control relations. But the assessment of physical and perceptual control is a highly embodied enterprise. Most of the time, knowing what one's interlocutor has physical or perceptual control over requires being in fairly rich perceptual contact with them. There are exceptions, of course. But the point is that the usefulness of demonstrative reference drops drastically when this sort of rich perceptual contact is blocked, even when both interlocutors know precisely where the other is located, where the conversational deictic center is. Data from naturally occurring communicative contexts of this sort (from sniper teams, and competitive online gaming) are presented which support this analysis.

Data:
1a. If you thought that/*this cookie was good, try this/*that.
1b. If you thought that/*this was a crazy idea, get a load of this/*that.

2a. Take this/*that wrench.                     2b. Listen to this/*that joke.
2c. Give me that/*this wrench.               2d. Tell me that/*this joke.

For a profile of Rick Grush, and more about his work, please see:

Rick's own website

Rick on the Department of Philosophy page

27 October - John Alderete, Department of Linguistics, Simon Fraser University and Paul Tupper, Department of Mathematics, Simon Fraser University

Integrating connectionist and symbolic approaches to phonotactics:
Arabic root cooccurrence restriction

In this talk, we develop a connectionist model of learning phonotactics and apply it to the problem of learning root cooccurrence restrictions in Arabic. In particular, a multilayer network with a hidden layer is trained on a representative sample of actual Arabic roots using error-corrective backpropagation. The trained network is shown to classify actual and novel Arabic roots in ways that are qualitatively parallel to psycholinguistic studies of Arabic.

Statistical analysis of network performance also shows that activations of nodes in the hidden layer correspond well with violations of symbolic well-formedness constraints familiar from generative phonology. The larger finding is therefore that a sub-symbolic phonotactic system trained from scratch can mirror the behavior of a symbolic-computational system typical of contemporary phonological analysis.

For a profile of John Alderete and of Paul Tupper, and more about their work, please see:

John's Linguistics Department biography
John's own website

Paul's Mathematics Department page
Paul's own website

17 November - Anoop Sarkar, School of Computing Science, Simon Fraser University

Learning by Bootstrapping

How can machines learn to process natural languages? In this talk we look at how machines might be able to learn to do some natural language processing tasks. We focus on bootstrapping methods which start from a small set of "seed" rules and iteratively learn new rules in an unsupervised manner, without any help from human experts.

We will look at three different language processing tasks: (1) disambiguating different senses of a word, (2) finding entities like persons, locations and organizations in text, and (3) the more complex task of machine translation from one language to another. Experimental evidence from these tasks show that bootstrapping is surprisingly effective, especially when compared to other unsupervised machine learning methods.

For a profile of Anoop Sarkar, and more about his work, please see:

Anoop's School of Computing Science page
Anoop's own website

23 June - Veronica Dahl, Professor, Computing, Simon Fraser University

Aiming at the Cognitive Web as a True Extension of the Human Brain

AI, while having given the world impressive accomplishments in specialized subfields, still falls quite short- other than partially and for very punctual systems or applications -of its ambitious original objective of endowing computers with human-like reasoning skills, including the ability to make useful sense of human languages. Yet our need for such skills have never been greater. For instance, biologists needing to determine if two proteins interact, direly need computers to read for them and “understand” enough of the vast number of written texts which mention them, in order to determine whether and which interactions might exist.

However in general, we cannot even reliably ask a computer simple questions, such as “Who won the Killam prize last year?”: whether with a keyword-based system or with a linguistically informed one, the best “answer” we can hope for is a deluge of documents that the user must wade through, and which may or may not contain the answer. In this talk we examine the historic reasons for this failure: lack of integration between the relevant disciplines involved, wasted efforts in reinventing the wheel, underutilization of NLU techniques at out disposal.

We maintain that each of the communities involved has

something that the others do not have and that should be shared for mutual benefit, and we observe that some of that sharing is starting to happen. Finally, we argue that for bringing computers closer to becoming true extensions of the human brain, we need to endow them with a cognitively-informed web, by integrating still underexploited connections and new methodologies around the pivot of Logic Programming, into an executable theory of machine-informed human cognition.

For a profile of Veronica Dahl, and more about her work, please see:

Veronica's webpage

10 February - Kathleen Akins, Department of Philosophy, Simon Fraser University

What is Neurophilosophy?

One very Canadian contribution to contemporary philosophical landscape is neurophilosophy. This is a branch of philosophy that uses current research from the neurosciences to answer —or at least re-direct inquiry about—a wide variety of philosophical/scientific questions. These might be specific questions about the nature of human mind (about the nature of self, mental representation, emotions, consciousness, our cognitive abilities and sensory processes) or even broader metaphysical questions about free will, the relation of the mind to the world in general, or about the utility of our scientific theories to understanding the human condition.

Given the rather, omnivorous nature of neurophilosophy, I would like to present some recent work on colour-grapheme synaesthesia. This makes a nice case study to demonstrate the “ball of string” nature of neurophilosophical questions. One starts with an empirical mystery, or an anomaly, or a philosophical view that seems suspect in light of some current neuroscientific evidence. In this case, we started with a rare and seemingly self-contained phenomenon—the fact that some people can see/think of letters as having specific colours. As you have all guessed, synaesthesia has lead us on a merry chase — through neurophysiology, psychophysics, developmental theories of cognition and memory, the nature of visual processing, colour vision, philosophical conundrums about qualia and consciousness in general. In this short talk, I will try to give some sense of where we are now, and in the process, some insight into what doing neurophilosophy is all about.

For a profile of Kathleen Akins, and more about her work, please see:

Kathleen's Department of Philosophy page
Kathleen's own website

17 March - Philippe Pasquier, School for Interactive Arts and Technology (SIAT), Simon Fraser University, Surrey

Agents, Multiagent Systems and Metacreation

Philippe Pasquier is pursuing both a scientific and an artistic research agenda. This presentation aims to introduce his past and current scientific works. Thus, this will be a wide overview rather than a focused, in depth, presentation. The (pragmatic and prosaic) goal is to inform the attendees (whether they are research students or faculty) and trigger some potential topics for further interactions and/or collaborations.

In a first part, scientific research projects conducted in the field of autonomous agents and multi-agent systems (MAS) will be introduced, with an emphasis on recent work on human behavior in computer-mediated interest-based negotiation. In a second part, various recent realizations in the field of computational creativity will be presented and discussed.

Dr. Pasquier's Website

School for Interactive Arts and Technology

After studying computer science, artificial intelligence and cognitive sciences in Europe, Canada and Australia, Philippe Pasquier joined Simon Fraser University's School of Interactive Arts and Technology in January 2008 as an assistant professor. Philippe Pasquier is both a scientist specialized in artificial intelligence and a multi-disciplinary artist interested in studying the links between Art, Science and Technology. His work has been shown on four continents and funded and supported by more than 20 scientific and or cultural institutions including the National Sciences and Engineering Research Council, the Canadian Council for the Arts, the French Ministère de la Culture et de la Communication, the Australian Research Council and the Australian Council for the Arts.

31 March - Phil Hanson, Department of Philosophy, Simon Fraser University

Topic: Epistemic Blindspots

By an 'epistemic blindspot' I shall mean a proposition that is true, but that cannot (humanly) be known to be true, by anyone, at any time.  Some key unclarities in this definition, due to the presence of such 'philosophically loaded' notions as "true", "cannot", and "known" will be addressed.  I will briefly sample early modern and contemporary reasons for supposing that there are, or are not, epistemic blindspots, most of which draw on considerations broadly construable as from the domain of "cognitive science".  The main focus of the talk is a famous little theorem whose formal proof , which is easily grasped, was published by the logician Frederick Fitch in the Journal of Symbolic Logic, in 1963.   (It has been referred to as "Fitch's Paradox" but usually only by those who think that there is something wrong with it.  It has also been referred to as the "Fitch/Church result" for reasons I will explain.)  Fitch's Theorem has the following conditional form: if there are truths that are (were, and will be) unknown by anybody, then there are also some unknowable truths. It has seemed utterly plausible to many that the antecedent of the conditional holds.  Modus Ponens then yields the consequent of the conditional, i.e., the existence of epistemic blindspots.  Others, though, have felt strongly enough that there are not any epistemic blindspots to bite the bullet and accept, via Modus Tollens and Fitch's Theorem, the denial of the existence of any unknown truths.  Still others have challenged the (minimal) assumptions on which the proof of the theorem is based.  The result is an active literature, still growing, some of it quite technical. In the brief time allotted I will review a few highlights and make brief observations about what Fitch's Theorem shows and doesn't show.

Dr. Hanson's website

21.Oct - Mark Blair, Department of Cognitive Science, Simon Fraser University

Why the Eye? Using Eye-tracking to Understand the Interplay Between Attention and Learning.

The latest generation of eye-tracking technology allows accurate, noninvasive recording of the allocation of overt visual attention. This technology is used widely to meet a diverse set of challenges in clinical, basic and applied settings. In the Cognitive Science Lab here at SFU, eye-tracking is used to study how people learn to pay attention to the important stuff in our world.

See Dr. Blair's website at SFU

30.Sept - Andrew Coward, Department of Computer Science, Australian National University, Canberra

The Hippocampal System as the Cortical Resource Manager: A model connecting psychology, anatomy and physiology

A model is described in which the hippocampal system functions as resource manager for the neocortex. This model is developed from an architectural concept for the brain as a whole within which the receptive fields of neocortical columns can gradually increase but with some limited exceptions tend not to decrease. The definition process for receptive fields is constrained so that they overlap as little as possible, and change as little as possible, but at least a minimum number of columns detect their fields within every sensory input state. Below this minimum, the receptive fields of some columns are increased slightly until the minimum level is reached. The columns in which this increase occurs are selected by a competitive process in the hippocampal system that identifies those in which only a relatively small increase is required, and sends signals to those columns that trigger the increase. These increases in receptive fields are the information record that forms the declarative memory of the input state. Episodic memory activates a set of columns in which receptive fields increased simultaneously at some point in the past, and the hippocampal system is therefore the appropriate source for information guiding access to such memories. Semantic memory associates columns that are often active (with or without increases in receptive fields) simultaneously. Initially, the hippocampus can guide access to such memories on the basis of initial information recording, but to avoid corruption of the information needed for ongoing resource management, access control shifts to other parts of the neocortex. The roles of the mammillary bodies, amygdala and anterior thalamic nucleus can be understood as modulating information recording in accordance with various behavioral priorities. During sleep, provisional physical connectivity is created that supports receptive field increases in the subsequent wake period, but previously created memories are not affected. This model matches a wide range of neuropsychological observation better than alternative hippocampal models. The information mechanisms required by the model are consistent with known brain anatomy and neuron physiology.

See Dr. Coward's website at Australia National University

18.Nov - Ron Rensink, Department of Cognitive Systems, UBC

Topic and details TBA

See Dr. Rensink's website at UBC

21.Jan - Joseph Stemberger, Department of Linguistics, UBC

Storing information in the human language system: Item-specific vs. system-general

There are two ways to store information in a system. (1) Store the information in a "ready-to-use" form in a specific stored item (such as a word). (2) Construct the information from more basic information, using more general principles. These are not mutually exclusive; partial information can be stored, and additional information can be added using more general principles. Debates (well, more like differences of opinion) regarding the nature of information storage lie at the centre of some of the hottest disagreements in research on human language today. I will explore this issue with reference to two aspects of human language production: phonology (the pronunciation of words) and morphology (the study of words with two or more meaningful parts, such as dog and -s of dogs). As always, the issue of what information is stored is difficult to disentangle from the issue of how the information is processed. Regardless of how much information is stored in particular items, there is also a general system that extracts that information in a way that affects the speed and accuracy of the processing of individual items; models differ on the nature of that general system. Error-driven learning mechanisms (where learning occurs primarily to correct errors that have occurred) lead to greater (or lesser) reliance on item-specific storage as a function of greater (or lesser) occurrence of errors during language production. I contrast approaches in which item-specific vs. system-general elements are competitors, and approaches in which they cooperate.

See Dr. Stemberger's website at UBC Linguistics.

Watch a video of Dr. Stemberger's talk

Dr. Stemberger is primarily interested in phonology, morphology, and their interaction. Most of his research is in adult psycholinguistics and in first language acquisition, and investigates the nature of cognitive representations. The adult work focuses primarily on errors during speaking, employing a variety of experimental tasks such as SLIPs (for phonological errors), and morphonaming and sentence production (for morphological errors), with some residual work involving his corpus of 7500 errors collected from spontaneous speech. Research on English child language (often in collaboration with May Bernhardt) involves collecting data from both normally-developing children and children with phonological delays/disorders, ChiLDES-based research, and analysis of diary-study data. He is also involved in a study of first language acquisition by monolingual Zapotec-learning children in San Lucas Quiavini, Oaxaca, Mexico, and am intending to begin a similar study on Slovene. His theoretical orientations are towards Optimality Theory, employing our own version of the theory, and towards connectionist models. Current research tends to focus on the degree to which phonological and morphological information is concentrated in particular lexical items vs. the degree to which lexical processing is supplemented with system-general information, but the acquisition projects look at all aspects of phonology and morphology.

18.Feb - Urs Ribary, Department of Psychology, SFU

Brain Network Oscillations and Human Perception

There is increasing evidence that human cognitive functions can be addressed from a robust neuroscience perspective. In particular, the distributed coherent electrical properties of central neuronal ensembles are considered to be a promising avenue of inquiry concerning global brain functions. Research in neuroscience further indicates that oscillatory activity in the gamma band (25-50 Hz) can be correlated with both sensory acquisition and premotor planning, which are non-continuous functions in the time domain. From this perspective, gamma band activity is viewed as serving a broad temporal binding function, where single-cell oscillators and the conduction time of the intervening pathways support large multicellular thalamo-cortical resonance that is closely linked with cognition and subjective experience. The working hypothesis is that although dedicated units achieve sensory processing, the cognitive binding process is a common mechanism across modalities. Moreover, it is proposed that such time-dependent binding when altered, will result in modifications of the sensory motor integration that will affect and impair cognition and conscious perception.

See Dr. Ribary's website at SFU Psychology.

Urs Ribary received his doctorate degree in Neuroscience and Neuropharmacology at the Swiss Federal Institute of Technology (ETH) in Zurich (Switzerland). During the past 19 years, he was holding a faculty position at New York University Medical Center (USA) and has served as the Director of the Center for Neuromagnetism, a functional brain imaging center, for 16 years. In 2007, Urs was awarded the endowed BC LEEF Leadership Chair in Cognitive Neuroscience in Childhood Health and Development and a Professor at SFU. Urs's goal is to coordinate multidisciplinary brain imaging resources development towards multi-disciplinary cognitive neuroscience brain imaging research across SFU, the Down Syndrome Research Foundation, the BC's Children Hospital and University of British Columbia. The focus will be to better understand the developing human brain, its cognitive abnormalities and pathologies and to develop better objective diagnostic procedures in relation to therapeutic interventions.

18.Mar - Hélène Gauchou, Department of Psychology, UBC

Tuning in on Zombies: When Implicit Cognition Turns Explicit

What does awareness do for cognition? Is it necessary to be conscious of an event in order to perceive it, remember it, or incorporate it in our abstracted representation? These questions about the relationship between performance and our mental state during performance have been part of psychology for many years. Dr. Gauchou will discuss a variation of those questions: Is it necessary to perceive an event in order to be aware of it? It has often been assumed that when we use vision to become aware of an event, this must be accompanied by a corresponding visual experience (i.e., seeing). But the existence of a particular effect, the Mindsight phenomenon, shows that this assumption is incorrect. When observers view a sequence of displays alternating between an image of a scene and a modified version of the same image, they often feel the change even though they have no visual experience of it. Dr. Gauchou will also address the problem of supposed-to-be-forgotten information, a classical example of implicit memory, but will introduce a new methodology still under construction — yet funny and fascinating, as it requires the use of a Ouija board.

Hélène Gauchou is a Post Doctoral fellow in the department of Psychology at the University of British Columbia. She works with Ron Rensink (Visual Cognition Lab, UBC) on different aspects of implicit cognition and among them the Mindsight phenomenon. She graduated in Cognitive Psychology and in 2007 obtained her PhD in Cognitive Science from Paris 5 / Paris 11 University. Her PhD research was conducted under the supervision of Kevin O'Regan, and was centred on the understanding of the nature and organisation of information stored in visual short term memory. She studied the mechanisms underlying change detection and assessed the existence of a new kind of information in memory, the relational information. She started a collaboration with a situated perception group (LIMSI - CNRS) in order to apply the notion of relational information to computer vision systems. She presented her work at major conferences on vision science (ECVP, VSS) and consciousness (ASSC). Her interests cover themes such as: visual perception, short term memory, implicit cognition, perceptual consciousness, and their interaction with attentional process.

1.Oct - Steve DiPaola, School of Interactive Arts and Technology

Computationally modeling artistic creative process using cognitive and AI approaches

Can you use computational intelligence techniques to allow aspects of art (the creative human act of fine art painting) and science (cognition, vision and perception) to both enhance and validate each other? This research takes two novel approaches to computer painterly rendering which rely on parameterizing a semantic knowledge space of how a human painter paints (the creative and cognitive process) as well as evolving programs that use techniques gleaned from current human creativity theory. These approaches have two intertwining and interdisciplinary research outcomes. The first is creating a new type of painterly rendering system with both a wider range and improved results compared to current techniques. The second benefit, is that portrait artists over 1000's of years have somewhat intuitively evolved a 'painting methodology' which exploits specific human vision and cognitive functions, and therefore when presented in a quantitative way (from our system) can shed light on psychological research in human vision and perception (or at least validate it via another method). The reverse is also true - via this system and process, cognitive scientists can understand artistic technique (which can be useful in many areas including how to make design systems creative). With this in mind we are 1) validating our system by using an original hypothesis about Rembrandt's use of cognitive textural agency in his late portraits as a test case and 2) using a Genetic Programming system to evolve portrait painting programs which use a human creativity based process. See http://ivizlab.sfu.ca/research for details.

Read an article about Steve's talk

Watch a video of Dr. DiPaola's talk

15.Oct - Sidney Fels, Electrical and Computer Engineering, UBC

Research at MAGIC and the Human Communication Technologies Lab

Communication and computing technology is advancing at an accelerated pace. Humans are finding it difficult to keep pace with these changes, and yet these new technologies are supposedly made for the benefit of humans. MAGIC and the Human Communication Technology Research Laboratory look at a number of key issues which put people "back in the loop" and allow us to communicate experiences to computer systems and each other more effectively. An awareness of people's different cognitive, physical and emotional capabilities provides a foundation for acquiring, analyzing, representing, storing, retrieving, transmitting, communicating and ultimately synthesizing human experience. Faster processing machines, bigger data capacity, new algorithms, multimedia and multimodal systems will be combined with developments in psychology, sociology and art to enhance the communication abilities between people and machines.

Visit Dr. Fels' website

Read an article about Dr. Fels' talk

Watch a video of Dr. Fels' talk

29.Oct - Tony Herdman, Department of Psychology

A love affair with MEG

Meg can captivate many with her looks and pleasant disposition. Her charm can entice the wisest explorer, but she is not for the faint hearted. Meg is complex and understanding her takes patience and perseverance. Dr. Herdman will describe some strengths and weaknesses of MEG's (magnetoencephalography's) ability to show us the beauty of our true love, the human brain. He will provide examples of how MEG can be used to investigate the neural functions underlying basic perceptual phenomena, motor execution, language, and memory.

My research programme explores the neural networks that underlie basic perceptual phenomena, motor execution, language, and memory. I use behavioural, eye-tracking, and electrophysiological (EEG & MEG) to measure such processes. My lab is currently focusing on three main research streams: 1) imaging the neural substrates of auditory, visual, audiovisual perceptions as they relate to reading and its typical and atypical development; 2) identifying electrophysiological evidence for the impact of socioeconomic disparity on auditory and visual attention in children and adults; 3) examining the interplay within cortical networks underlying eye movements that are indicators of memory consolidation and retrieval.

Find out more at the Human Brain Research Lab's website

Read an article about Dr. Herdman's talk

Watch a video of Dr. Herdman's talk

12.Nov - John Alderete, Department of Linguistics

Tackling language learnability problems with some core ideas from Optimality Theory

This presentation summarizes a body of work that shows how certain theoretical assumptions about the nature of grammar in Optimality Theory (OT, Prince & Smolensky 1993/2004) can lead to tractable solutions to complex learning problems. In OT, grammars are orderings of well-formedness constraints that produce the desired input-output mappings, and the output-oriented nature of this theory requires a kind of evaluation mechanism that compares every winning form with dozens of logical losers. An error-driven model of grammar learning is shown to work with these assumptions to efficiently learn complex linguistic systems from overt language forms, tackling several difficult learning problems, including efficient search of the space of possible grammars, inferring hidden structure and underlying lexical representations when the correct phonological grammar is not known, and deducing how paradigmatic information informs lexical representations. Conversely, the learning model also contributes to linguistic theory: certain unattested surface alternations can be shown to be unlearnable in this model, so the lack of such patterns typologically could be explained by an external learnability filter.

Visit Dr. Alderete's website

Read an article about Dr. Alderete's talk

Watch a video of Dr. Alderete's talk

26.Nov - Lyn Bartram, School of Interactive Arts and Technology

Whisper, Don't Scream: Design, Vision and Visualization (or, More Eyes on the Prize)

The utility of visualization to a wide variety of tasks and domains has been well established over years of research and is increasingly grounded in studies of human perception and attention. Much of this research has been concerned with what are the most salient representations (i.e. those that attract and retain attention in perceptually efficient ways). But as the scope of the information to be displayed grows increasingly complex and the ways in which we use visualization proliferate, we are faced with a new challenge: how do we integrate and balance demanding visual cues such that they are appropriately relevant without screaming for attention? Moreover, how do we adapt or extend our notions of representation to gracefully incorporate new scopes of meaning, in particular affect and emotion, that do not necessarily function on highly focused attention but rather on ambient or peripheral communication? The disciplines of visual design, art and psychology are essential to characterising solutions to these questions. In this talk I will discuss several current research projects that address these issues and discuss the need for an integrated approach.

Visit Dr. Bartram's website

Read an article about Dr. Bartram's talk

Watch a video of Dr. Bartram's talk

16.Jan - Fred Popowich, School of Computing Science

Does My Computer Really Understand Me?

How can computers be used to process human language, either to make it easier for human beings to interact with computers, or to make it easier for human beings to interact with each other? Typical real world applications of this research include the automatic translation of language, tools to assist people in learning language, and tools to help people search and manage the information contained on computer systems. During his presentation, Dr. Popowich will examine how cognitive science research is now being used in information technology systems, looking at the benefits and challenges associated with language technologies. He will also look at how these technologies might be used in the future.

Read the article about Dr. Popowich's talk

30.Jan - UNIVERSITY CLOSURE

13.Feb - Tracey Leacock, Faculty of Education

Exploring Technology-mediated Writing and Learning

Dr. Leacock’s primary research interest is investigating how computer technologies can be used to help students make better judgments and decisions about their writing, and how writing can help students to learn. Writing has long been recognized as an area in which students need to do
well - but don’t. How do teaching methods - such as using detailed scoring rubrics and providing cognitive tools to help students revise - affect students’ abilities to evaluate their learning and adapt their studying? This work spans the fields of educational technology, self-regulated learning, and academic literacy. An important consideration in Dr. Leacock’s work is to build a solid
foundation of knowledge and technological tools that can be used in ongoing work to investigate writing in technology-mediated environments.

27.Feb - Brian Fisher, School of Interactive Arts and Technology

Visual Analytics: Science & Application

Visual Analytics is the science of analytical reasoning supported by the highly interactive visual interface. People apply visual analytics research to guide the design of tools and techniques that augment human cognitive processing: to derive insight from massive, dynamic, and often conflicting data and to provide timely, defensible, and understandable assessments of complex situations. Visual analytics applications combine scientific and information visualization which include statistics, mathematics, cognitive and perceptual sciences and decision sciences. Application areas include scientific inquiry, the planning and coordination of disaster relief, law enforcement, business intelligence, risk analysis, general investigative analysis, operations management, and a range of other areas.

See also: Vancouver Conference and Exploratory Workshop held, jointly with the University of British Columbia, and Simon Fraser University

12.Mar - John Alderete, Department of Linguistics

Tackling language learnability problems with some core ideas from Optimality Theory

** POSTPONED. CHECK BACK FOR RESCHEDULED DATE **

This presentation summarizes a body of work that shows how certain theoretical assumptions about the nature of grammar in Optimality Theory (OT, Prince & Smolensky 1993/2004) can lead to tractable solutions to complex learning problems. In OT, grammars are orderings of well-formedness constraints that produce the desired input-output mappings, and the output-oriented nature of this theory requires a kind of evaluation mechanism that compares every winning form with dozens of logical losers. An error-driven model of grammar learning is shown to work with these assumptions to efficiently learn complex linguistic systems from overt language forms, tackling several difficult learning problems, including efficient search of the space of possible grammars, inferring hidden structure and underlying lexical representations when the correct phonological grammar is not known, and deducing how paradigmatic information informs lexical representations. Conversely, the learning model also contributes to linguistic theory: certain unattested surface alternations can be shown to be unlearnable in this model, so the lack of such patterns typologically could be explained by an external learnability filter.

26.Mar - Nancy Hedberg, Department of Linguistics

Cognitive Status and the Form of Referring Expressions

Speakers select the form of their referring expressions according to their estimation of the status of the referent in the mind of the hearer.  For example, it is appropriate to say, "the dog next door kept me awake" when the hearer is not familiar with the dog, but saying, "that dog next door kept me awake" requires that the hearer already be familiar with the dog. To say "this dog kept me awake" requires, in addition, that the dog is in the current awareness of the hearer, at least with the help of a gesture. To say "It kept me awake" requires in addition that the dog be at the center of attention. This talk introduces the "Givenness Hierarchy", a theoretical construct which expresses the relationship between such cognitive status and their linguistic correlates, and discusses how it relates to issues in Cognitive Science.

26.Sept - Anoop Sarkar, School of Computing Science

Compute Less, Explain More: Computational Constraints on Linguistic Descriptions

The scientific study of language involves the use of various notational devices that are used to concisely describe rules that explain our linguistic behaviour. For example, a linguist may express a phonological rule in a certain language, that the correct pronunciation of a word involves vowel harmony, so that 'kikede' is ill-formed but 'kikidi' is well-formed. Or a linguist may express the syntactic rule that explains why 'Who did China invade?' or 'The stocks fell because China invaded Taiwan.' is acceptable but 'Who did the stocks fell because China invaded?' is not.

I will attempt to convince you that paying attention to computational constraints on these notational devices can help generalization and explain the often observed cross-linguistic "locality" of linguistic rules. By exploiting a computational perspective, these linguistic notational devices can also be shown as computable, either by brains or machines.

Read the article about Anoop's talk

10.Oct - Mark Blair, Cognitive Science Program and Department of Psychology

Information Emphasis and Access in Human Cognitive Systems: What medical diagnostics, bird-watching, and airport security have in common

People learn to attend to information that is helpful for distinguishing the categories being acquired, and to ignore information that is not. While the allocation of attention is thought to be dynamic, shifting with experience to reduce error, it is also assumed to be a top-down process that is independent of stimulus features. I will review recent studies in my lab which use eye-tracking technology to question this assumption and which reveal new insights into human selective attention.

Mark's basic research is focused on improving our understanding of human cognition generally. This knowledge can then be used by applied scientists for many practical purposes, such as designing automatic computer-based classification systems, improving training procedures for medical personnel, or for airport security personnel who need to detect weapons in the images on their luggage scanners. His work also has implications for how we learn regularities in language; for example, the parsing of sound into phonemes, the basic components of words.

Read the article about Mark's talk

24.Oct - Jim Delgrande, School of Computing Science

Belief Revision and Reasoning about Action

A cognitive agent, in a realistic scenario, will need to function in a dynamic domain.  Such an agent will make observations concerning the state of the world, carry out actions that change the state of the world, make and perhaps communicate with or observe other agents.  This talk will provide an (extremely) brief introduction to the related areas of belief change and reasoning about action and sketch some of the problems that arise.

7.Nov - Maite Taboada, Department of Linguistics

Thumbs Up or Thumbs Down? Detecting Sentiment and Opinion Automatically

The web is now the place to find opinions, ideas and new trends. As marketers, politicians and strategists already know, the success or failure of a "product", widely understood, is often dependent on what people say about it on-line. Part of my research focuses on how we can automatically extract sentiment. Research on opinion and subjectivity in text has grown considerably in the last few years. New methods are being created to distinguish objective from subjective statements in a text, and to determine whether the subjective statements are positive or negative with respect to the particular subject matter. In this talk, I will describe current work that aims at determining the sentiment of a text by examining its discourse structure.

Read the article about Maite's talk....

21.Nov - Oliver Schulte, School of Computing Science

Automated Scientific Discovery in Particle Physics

I will present some of my work on automated scientific discovery in particle physics. This work combines both engineering/computer science aspects and cognitive modeling aspects; I use it as a starting point to reflect on the vision of cognitive science laid out in Newell and Simon's famous Turing award lecture (1975).

An important problem in particle physics has been the search for laws that govern the dynamics of subatomic particles. This search has led to the discovery of new conservation laws (e.g., Baryon charge) to join classical conservation laws (e.g., electric charge, momentum). I describe a machine learning program that given current reaction data from particle accelerators, rediscovers the same conservation laws that physicists have found. The program is based entirely on general machine learning principles, such as Computational Learning Theory and the Minimum Description Length (MDL) principle. It also turns out to be a good model of the inferences of particle physicists.
I take this happy coincidence as an occasion to reflect on Newell and Simon's vision of what the division of labour should be between the engineering/normative disciplines in cognitive science (CS/AI, logic, rational choice theory) and the experimental/descriptive disciplines (psychology, linguistics). Have things turned as Newell and Simon expected they would in 1975?

Read the article about Oliver's talk...

17.Jan - Bill Turnbull

Talk-as-social-interaction

The universal and frequent activity of two or more persons talking to one another is constituted by the sequential alternation of conditionally relevant turns. Thus, talk is a universal and frequent form of social interaction. Contrary to dominant views, talk is not a disorderly and degraded reflection of covert linguistic competence. As documented by research in Conversation Analysis, talk is an orderly activity that participants achieve together through the use of orderly procedures or practices. Those orderly practices are indexical, reflexive, (mainly) sequential, observable in the details of talk, co-constructed, defeasible, tacit (in certain aspects), moral, and oriented to by participants. The practices for doing talk are not descriptions of the causes of social interaction but rather are templates in terms of which contributions to interaction are found to be recognizably intelligible, or not. Further, although the practices are discovered empirically (in Conversation Analytic research), the practices themselves are a priori objects given to us by the culture. In the presentation, I explore implications for the cognitive sciences of a model of talk-as-social-interaction.

31.Jan - John McDonald

Dynamics of attentional control in the human brain

Neuroimaging has identified brain areas that control voluntary shifts of attention, but the dynamics of attentional control are largely unknown because the hemodynamic response is too slow to track brain activity across time. We investigate the dynamics of attentional control by mathematically reconstructing the neural sources of electroencephalographic activity during a attention-cuing tasks. We looking at sources of low-frequency (theta) activity, we found pccipital, parietal, and frontal cortices were activated in succession within 500 milliseconds of a signal to shift attention. Parietal and occipital areas were then re- activated in succession prior to the appearance of a visual target. These results show that attentional control involves a feed-forward projection to frontal cortex and a subsequent feed-back projection to visual cortex, largely through the same cortical regions.

14.Feb - Richard Vaughn

Careful Coupling of Autonomous Systems: Practical robot-robot and robot-animal interaction

This talk considers the design of robot-robot, robot-animal, and robot-human interactions from an interface point of view. We consider that an intelligent autonomous system is striving to achieve some goals, whether survival and reproduction in animals, or work in robots. Agent-agent interfacing is characterized as a coupling of the dynamics of both parties; effective interfacing affords progress by both parties towards their goals. Some examples of real robot-something interaction will be presented.

28.Feb - Chung-hye Han

An experimental investigation into the placement of the verb in Korean and Japanese

In a head final language, the placement of the verb is hard to detect, since there is no evidence from the string to distinguish a verb-raising analysis from a a non-verb-raising analysis. This is so both for children acquiring the language and for linguists developing an analysis of it. Could the lack of input data lead to multiple grammars in a single speech community when it comes to verb placement? In this talk, I present evidence supporting this expectation, using experimental data concerning the scope of quantified noun phrases (NPs) and negation in Korean and Japanese. The results from this work is consistent with recent proposals in diachronic syntax and language acquisition literature that even given the restricted hypothesis space determined by Universal Grammar (UG), insufficient input can lead to distinct grammars in a single population.

8.Nov - Christine MacKenzie

Hand and Mind: From Prehension to Comprehension

What do hands do? Look around you for examples of the power of the human hand: the construction of buildings, the grooming of our bodies, the design of furniture, the intricacy of tool design...

22.Nov - Brian Fisher

A Cognitive Science Approach to the Design of Information Visulaization Systems