Colloquium Schedule

Institute of Cognitive Science Colloquium Schedule

All colloquia take place on Friday from 12:00 to 2:00 pm (unless otherwise noted) in Room D428 and D430 on the fourth floor of the Muenzinger Psychology Building.

If you have questions, please call the ICS Office at 303-492-5063.

Spring 2016

January 15, 2016 ~ Opening Session
Marie T. Banich
Director, Institute of Cognitive Science, University of Colorado Boulder
Title:  Drug Abuse and the Brain

Abstract: In this talk, I will discuss prior work that I have performed with colleagues at CU Denver on alterations in brain systems found in youth and adults who abuse substances. Following that I will discuss a new national project, the Adolescent Brain Cognitive Development (ABCD) project, which is designed to examine the development of brain systems related to substance abuse, including those responsible for cognitive control and reward processing. This project will also involve collecting a host of other data on cognitive, emotional, and social development. The project will be longitudinal in nature, spanning a 10-year time period, with data collected on over 10,000 individuals at close to 20 sites across the country. CU Boulder has been selected as one of the institutions to be involved in this landmark study, and will be doing so through a joint collaboration between the Institute of Cognitive Science and Institute for Behavioral Genetics. In particular, we, along with three other institutions, will be collecting data on twin pairs to disentangle genetic versus environmental contributions to drug abuse. The on-going design and implications of this study will be discussed.

January 18, 2016
Martin Luther King Day, Campus

January 22, 2016
Christine Brennan

Assistant Professor, Speech Language and Hearing Sciences, University of Colorado Boulder

Title: Phonological grain size sensitivity in auditory cortex is related to reading skill

Abstract: Previous evidence reveals regions within the auditory cortex selectively activate based on phonological grain size (number of speech sounds presented) (DeWitt and Rauschecker, 2012). Here, we employed direct testing of phonological grain size in order to confirm and further delineate this organization. While reading impairment is associated with deficits in phonological skill, including difficulty isolating phonemes (Shaywitz et al., 1998; Temple et al., 2001), it is unknown if selective activation for grain size is related to reading skill. Since isolation of small grain units is problematic in dyslexia, reading skill may be related to selectivity of the auditory cortex, especially for small grain stimuli. In this study, we examined the relationship between grain size selectivity and reading skill. Using functional magnetic resonance imaging (fMRI), we studied 20 typical adults under different grain size conditions. Stimuli included speech with one, two, or four phonemes. Stimuli were presented in blocks of seven trials from the same condition. During scanning, an active listening task was completed to ensure participants were awake and attending to stimuli. A timed measure of reading skill (TOWRE) was completed outside the scanner (Torgesen et al., 1999). Using SPM8, we contrasted the phonological grain size conditions. We also completed ROI analyses to examined brain-behavior correlations between activation strength and reading skill. Specifically, we compared the difference in activation for the small versus the large conditions with reading skill. We found significantly greater activation in bilateral middle superior temporal gyrus (m-STG) for 1-2 speech sounds (small grain size) and greater activation in the left middle temporal gyrus (MTG), right anterior STG, and right posterior STG for 4 speech sounds (large grain size). Brain-behavior analyses revealed significant correlation of reading skill with activation for the small compared to large grain size condition (p < .01). Adults who had higher reading skill activated the mid-STG more than lower skill readers. In addition, adults with higher skill engaged the left MTG less than lower skill readers for the large compared to small grain size condition, although this trend was not statistically significant. These results further delineate the organization of the temporal cortex, revealing not only selective activation related to phonological grain size, but also that selectivity is related to reading skill. Higher reading skill was associated with stronger engagement of the m-STG, suggesting that sensitivity to single phonemes may be linked to better reading. Importantly, these results provide a link between phonological grain size sensitivity in the brain and reading skill, consistent with behavioral studies that have shown strong correlations between phoneme awareness and reading ability. The findings have implications for developmental dyslexia, a condition often associated with deficits in phonology. Future studies should investigate phonological organization in children with and without dyslexia to determine if differences in grain size representation underlie the disability.

January 29, 2016
Thomas Hills
Professor, Department of Psychology, University of Warwick, UK

Title: Control and Representation in Cognitive Search 

Abstract: The trade-off between exploration and exploitation is a ubiquitous feature of animal life. Neuromolecular and behavioral evidence from across species suggest the ability to mediate this trade-off originated approximately 700 million years ago in a spatial foraging behavior called area-restricted search, allowing animals to modulate foraging behavior in response to resource density. Further evidence suggests this architecture was later exapted in vertebrates to modulate attention and search in internal representations: to maintain goals in the absence of external stimuli and to look, so to speak, before we leapt. In this talk, I will present research from my lab investigating internal search using comparisons of computational models that combine quantitative representations of internal environments with control processes that can navigate these environments. The representations are derived from multiple sources, including unsupervised learning from natural language corpora, social networks, and problem representations based on solution similarity. The control processes consist of random walks and multi-stage models that include dynamic transitions between representations. Over a series of studies, this work suggests that internal search is a form of area-restricted search consistent with the principles of optimal foraging in space and is governed by individual differences in executive control. This approach offers insights into cognitive control not offered by standard proof-of-principle approaches and helps to develop both an evolutionary and process-based account of cognitive control. I will conclude by presenting applications and questions posed by this work for age-related cognitive decline and changes in lexical representation across the lifespan.

February 5, 2016
Sidney D'Mello
Assistant Professor, Joint Appointments in Department of Computer Science and Engineering (CSE) and Department of Psychology, University of Notre Dame

Title: Between Boredom and Bewilderment: Coordinating Feeling and Thinking to Optimize Learning

Abstract: We study the complex interplay between cognitive and affective states (e.g., confusion, frustration, mind wandering) during learning and leverage insights to develop technologies that  coordinate what learners think and feel in addition to what they know and do. Our basic research investigates how complex mental states arise and influence learning via an analysis of interactions among the learners themselves, the learning content, and the learning activity. We then use signal processing and machine learning techniques to build computational models of mental states from facial features, body movements, peripheral physiology, eye gaze, and contextual cues in a variety of digital learning environments, both in the lab and in the wild. Finally, we close the loop by embedding our models in affect- and attention- aware technologies that increase engagement and learning by dynamically adapting to cognition and emotion. This talk will discuss our theoretical foundations, summarize key findings, and discuss our vision for the future of cyberlearning.

February 9, 2016 (Tuesday)
Steven Bethard
Assistant Professor, Department of Computer and Information Sciences, University of Alabama at Birmingham

Title: Parsing the Language of Time

Abstract: Getting a computer to understand the timeline underlying a written narrative is a critical component of tools for review of patient medical histories, analysis of intelligence reports, and tests of reading comprehension. But human language is rarely explicit in the way that would be most convenient for a computer, and events, times, and temporal relations are often implicit, left to be inferred by the reader. In this talk, I will first present a typical computational methodology for constructing timelines from the explicit and implicit cues of language: a series of supervised machine learning components trained on example texts whose timelines have been annotated manually by humans. Then I will show how we can improve this approach by analyzing big data that has not been annotated by humans but nonetheless reveals patterns in how humans talk about time. Finally, I will present an alternative approach to inferring timelines from text that achieves better generalization through modeling the incremental and compositional nature of the language of time.

February 19, 2016
Jeremy Reynolds
Senior Data Scientist Lead, Advanced Analytics and Data Science, Information Management & Machine Learning, Microsoft

Recording | Slides

Title: Stories from Industry: How can Cognitive Science Programs enable students to succeed?

Abstract: Aspiring graduate students often wish to develop their careers through the pursuit of tenure-track faculty positions. While tenure-track positions certainly have a number of benefits, they are a) not very accessible, and b) not for everyone, even if you have access. During this talk, I will discuss my experiences in both a tenure-track faculty position and industry, and I will discuss how strong training in cognitive science and experimental methods can prepare individuals for either type of career. I will also discuss the gaps I have seen in training programs, how students may want to augment their skills to fill those gaps before entering the job market, and how mentors can support students in pursuits outside of academia.

Tuesday, February 23, 2016
1:00-2:00 PM
*Talk sponsored by INC*
Cherie Marvel

Director, Cognitive Neurospsychiatric Research Laboratory (CNRLAB)
Dept. of Neurology, Div. of Cog. Neuro. Dept. of Psychiatry & Behav. Sciences Johns Hopkins School of Medicine

Recording

Title: Attention-to-reward as a predictor of HIV-related risky behaviors.

Abstract: Human immunodeficiency virus (HIV) is often contracted through reward-seeking, risky behaviors, such as needle sharing and unprotected sex. Understanding the factors that motivate an individual to engage in risky behaviors is important to limiting the spread of HIV. One's attention to reward, and conversely, one's ability to ignore that reward, may represent a potential source of vulnerability. In this talk, Dr. Marvel will present data that examines the link between attention-to-reward and HIV-related risk-taking behaviors in HIV positive individuals. 

February 26, 2016
Peter Foltz

Institute of Cognitive Science Research Professor, University of Colorado Boulder

Title: Automated Analyses of Language Production in Clinical Tasks

Abstract: Language provides a window into underlying cognitive structures and mechanisms. In disorders such as schizophrenia, psychosis and bipolar disorder, abnormalities in language production and comprehension are often used as indicators to aid in the diagnosis and in understanding the underlying etiology. This talk will describe research in which we apply automated language analysis techniques to detect differences between groups of control and clinical populations. Over the past 15 years, we have collected samples of language from patients, controls, and unaffected siblings performing a range of tasks in clinical trials and neuropsychological research studies. The tasks include category fluency, logical memory, story recall, story telling, and answering open-ended questions. The analyses use corpus-based statistical models of language to examine semantic and statistical properties of responses. The results indicate that the methods can reliably detect differences in regularities in language between groups. Implications will be discussed for using the approach as a framework for measuring subtle changes in language and understanding the underlying processes that may cause these changes. Finally, I’ll talk about applying the approach for remote monitoring and treatment of cognitive functioning on mobile devices.

March 4, 2016
Frank Jäkel
Assistant Professor for Cognitive Modeling, Institute of Cognitive Science, University of Osnabrück, Germany

Title: Categorization: From Psychology to Machine Learning and Back

Abstract: The ability to categorize is fundamental for cognition—in humans and in machines. Many, if not all, so-called higher cognitive functions, like language or problem-solving, crucially depend on categorization. For this reason, research on categorization plays a central role in cognitive science and artificial intelligence alike. Hence, it is no surprise that many successful machine learning algorithms for categorization were inspired by results and insights from psychology and neuroscience. However, today machine learning is a mature field and more recent methods are usually seen to be grounded in statistics and computer science rather than in cognitive science. Kernel methods, in particular, have gained popularity in machine learning and have proved to be successful in many applied categorization problems. I will describe how similar ideas have developed in psychology and how insights from machine learning can feed back into cognitive science.

March 11, 2016
Joshua Correll

Associate Professor, Department of Psychology and Neuroscience, University of Colorado Boulder

Title: Of Kith and Kin: Perceptual Enrichment, Expectancy, and Reciprocal Processing in Face Perception

Abstract: Race powerfully affects perceivers’ responses to faces, promoting biases in attention, classification, and memory. To account for these diverse effects, we propose a model that integrates social cognitive work with two prominent accounts of visual processing: perceptual learning and predictive coding. Our argument is that differential experience with a racial ingroup promotes both (a) perceptual enrichment, including richer, more well-integrated visual representations of ingroup relative to outgroup faces, and (b) expectancies that ingroup faces are more normative, which influence subsequent visual processing. We present evidence from three lines of research concerning predictions of the model.

March 18, 2016
John Trueswell

Professor of Psychology, Institute for Research in Cognitive Science, University of Pennsylvania

Title: The Role of Cognitive Flexibility in Language Processing and Language Acquisition

Abstract: Because children and adults interpret speech in real-time, rapidly making commitments to interpretation essentially on a word-by-word basis, they must learn to deal flexibly with temporary ambiguities that arise in the input. In this talk, I'll present a series of experiments that examine the relationship between language processing, language learning, and cognitive flexibility. It is found that individual differences in how well children flexibly respond to representational conflict during executive function tasks predict how well they deal with temporary syntactic ambiguity during real-time comprehension. I explore how the processing challenges associated with real-time comprehension constrain grammar learning, and might even constrain the types of grammars that arise in languages of the world. The results reveal some of the ways in which the cognitive abilities of the individual shape the linguistic-communicative system of the group.

March 21-25, 2016
Spring Break, Campus Closed March 25th

April 1, 2016
Derek Lomas
Design Fellow at the UC San Diego Design Lab

Title: Large-Scale Online Experiments Can Accelerate the Production of Interaction Design Theory

Abstract: Each day, companies run thousands of design experiments to optimize software designs (e.g., A/B tests). What if just a fraction of these experiments were used to test generalizable theories about how and why interaction designs affect user behavior? How would this impact the field of interaction design?

To illustrate the opportunities for online theory testing, I will share several experiments investigating a classic theory in game design, that games should be neither too hard or too easy. This has been formalized as the hypothesis that “moderate difficulty will produce optimal motivation”. To test this hypothesis, I deployed thousands of variations of an online educational game to >50,000 users. Surprisingly, the results showed that low levels of difficulty consistently produced maximum motivation. Further experiments indicated that the factor of "novelty" plays a critical and rarely recognized role in maintaining player motivation. These findings show how online experiments can be used to build generalizable theories about how designs impact users.

More broadly, these results illustrate how the proliferation of large-scale online theoretical experiments could rapidly produce a large body of empirically-validated interaction design theory. How can we prepare for a "big science" of interaction design? To explore this, I share some additional experimental work with "multi-armed bandits", which suggest how artificial intelligence might participate in the future of large-scale scientific inquiry. I conclude by discussing several of the opportunities, limitations and risks of scientific theory-making in the field of design.

April 8, 2016
Anu Sharma

Professor, Speech Language and Hearing Sciences, University of Colorado Boulder

Title: Cross-Modal Brain Changes in Hearing Loss Across the Age Spectrum

Abstract: A basic tenet of neuroplasticity is that the brain will re-organize following sensory deprivation. Auditory deprivation appears to tax the brain by changing its normal resource allocation. Compensation for the deleterious effects of hearing loss may include recruitment of alternative or additional brain networks to perform auditory tasks. Our high-density EEG experiments suggest that age-related hearing loss results in significant changes in neural resource allocation, reflecting patterns of increased listening effort, decreased cognitive reserve, which may be associated with dementia-related cognitive decline. Cross-modal plasticity is another form of cortical re-organization associated with deafness. Cross-modal plasticity occurs when an intact sensory modality recruits cortical resources from a deprived sensory modality to increase its processing capabilities as compensation for the effects of sensory deprivation. Our results suggest evidence of recruitment of higher-order auditory cortical areas by visual and somatosensory modalities in hearing loss and deafness. Cross-modal cortical re-organization is evident both in congenital deafness and in age-related mild-moderate hearing loss and shows a strong negative correlation with speech perception performance. Overall, our results suggest that compensatory cortical plasticity secondary to sensory deprivation has important neurological consequences and influences outcomes in children and adults with hearing loss.

April 22, 2016
Bob L. Sturm

Lecturer in Digital Media, School of Electronic Engineering and Computer Science, Queen Mary University of London

Title: Clever Hans, Clever Algorithms: Are Your Machine Learnings Learning What You Think?

Abstract: In machine learning, generalisation is the aim, and overfitting is the bane; but just because one avoids the latter does not guarantee the former. Of particular importance in some applications of machine learning is the “sanity" of the models learnt. In this talk I discuss one discipline in which model sanity is essential -- machine music listening — and how several hundreds of research publications may have unknowingly built, tuned, tested, compared and advertised “horses” instead of solutions. The true cautionary tale of the horse-genius Clever Hans provides the most appropriate illustration, but also ways forward.

April 29, 2016
ICS Poster Session and Mexican Fiesta, Last Day of Classes