Two Approaches to Analogy: Deep Learning versus Structural Models
When: Nov 17, 2021 2:00 PM Arizona
Speaker: Hongjing Lu - Departments of Psychology and Statistics, UCLA
Abstract: Human perception and reasoning goes beyond knowledge about individual entities, extending to inferences based on relations between entities. We recognize objects from visual input and understand meanings of entities from text input, and also see the relations among objects and entities in their context. It remains unclear what types of representations are deployed to achieve these feats. Can human perception and reasoning be best emulated by applying deep learning models to massive numbers of reasoning problems, or should learning instead focus on acquiring structural representations, coupled with the ability to compute similarities based on such representations? To address this question, I will present two modeling projects, on visual analogy and on verbal analogy. We compare human performance to predictions derived from deep learning models and from models based on structural representations. Our approach demonstrates that human-like analogical mapping can emerge from comparison mechanisms applied to rich semantic representations of individual concepts and relations. I will argue that perception, semantics and reasoning are closely coupled, and that structural representations play an essential role in facilitating analogical reasoning.
After registering, you will receive a confirmation email containing information about joining the meeting.
Going beyond the here and now: Counterfactual simulation in human cognition
Assistant Professor of Psychology,
October 20, 2021 2PM MST
As humans, we spend much of our time going beyond the here and now. We dwell on the past, long for the future, and ponder how things could have turned out differently. In this talk, Dr. Gerstenberg will argue that people's knowledge of the world is organized around causally structured mental models, and that much of human thought can be understood as cognitive operations over these mental models.
The Machine Learning Human Footprint Index
Date: Oct 6, 2021 02:00 PM in AZ
Abstract: The human footprint index (HFI) is an extensively used tool for interpreting the accelerating pressure of humanity on Earth. Up to now, the process of creating the HFI has required significant data and modeling, and updated versions of the index often lag the present day by many years. Here we introduce a near-present, global-scale machine learning-based HFI (ml-HFI) which is capable of routine update using satellite imagery alone. We present the most up-to-date map of the HFI, and document changes in human pressure during the past 20 years (2000–2019). Moreover, we demonstrate its utility as a monitoring tool for the United Nations Sustainable Development Goal 15 (SDG15), 'Life on Land', which aims to foster sustainable development while conserving biodiversity. Moving forward, the ml-HFI may be used for ongoing monitoring and evaluation support toward the twin goals of fostering a thriving society and global Earth system.
Bio: Dr. Barnes: Dr. Elizabeth (Libby) Barnes is an associate professor of Atmospheric Science at Colorado State University. She joined the CSU faculty in 2013 after obtaining dual B.S. degrees (Honors) in Physics and Mathematics from the University of Minnesota, obtaining her Ph.D. Dr. Barnes' research is largely focused on climate variability and change, and the data analysis tools used to understand it. She has received many prestigious awards including AGU Macelwane Medal, AGU Turco Lectureship and an NSF CAREER grant.
Dr. Keys: Dr. Patrick Keys is a Research Scientist for SoGES. His research is focused on a broad range of global sustainability challenges, including climate change impacts, cross-scale risks, and social-ecological tele-connections. Prior to joining SoGES, Pat founded an environmental consultancy that worked with local and international partners.
Critical Digital Studies, DH, and Ethical Collaboration
Liz Grumbach & Sarah Florini (ASU)
Critical digital studies and digital humanities share an investment in interrogating and disrupting systems of power and oppression. Yet, they are rarely in dialogue. In their talk, Liz Grumbach and Sarah Florini will discuss projects they are undertaking that put these disciplines in conversation. Undertaken in cooperation with the Lincoln Center for Applied Ethics under the broad umbrella of “Technologies of Domination,” these ongoing projects include a two-day design lab workshop on technology and oppression, a collaborative research endeavor with TikTok creators, and a de(con)structive approaches to whiteness and UX. Grumbach and Florini demonstrate how critical digital studies and digital humanities can form a generative partnership grounded in ethics of collaboration and community.
Liz Grumbach is the Program Manager for Digital Humanities and Research for the Lincoln Center for Applied Ethics, an organization committed to exploring co-creative and participatory strategies for ethical technological innovation, at Arizona State University. Her work has been published in scholarly publications such as Digital Humanities Quarterly, Scholarly Research and Communication, Digital Scholarship in the Humanities, and the Journal on Computing and Cultural Heritage.
Sarah Florini is an Associate Professor of Film and Media Studies in the Department of English at Arizona State University. Her work combines critical digital studies, critical race theory, and African American and Black Studies to explore the intersection of race and technology. Her book, Beyond Hashtags: Racial Politics and Black Digital Networks, is available on New York University Press.
Date: September 22, 2021
Time: 2:00PM - 3:00PM Arizona
Network Inference from Grouped Observations
Speaker: Dr. Yunpeng Zhao, School of Mathematical and Natural Sciences
Date: September 8, 2021
Time: 2:00PM - 3:00PM Arizona
Download flyer: b2c2-flyer.pdf
After registering, you will receive a confirmation email containing information about joining the meeting.
Abstract: Statistical network analysis typically deals with inference concerning various parameters of an observed network. In several applications, especially those from social sciences, behavioral information concerning groups of subjects are observed. Over the past century a number of descriptive statistics have been developed to infer network structure from such data. However, these measures lack a generating mechanism that links the inferred network structure to the observed groups. In this talk, we present a model-based approach called the hub model, which belongs to a family of Bernoulli mixture models. We further present theoretical results on model identifiability, a notoriously difficult problem in Bernoulli mixture models, and estimation consistency.
Bio: Yunpeng Zhao is an associate professor in the School of Mathematical and Natural Sciences in New College of Interdisciplinary Arts and Sciences at Arizona State University. His primary research interest includes machine learning methodology and theory in network analysis with applications in biology and the social sciences. He is also working on high dimensional data analysis with applications in genomics.
B2C2 Data Initiative Seminar - April 14, 2021
Title: Developing An App To Address Police Officer Stress & Decision-Making: Goals, Challenges, and Cross-Disciplinary Opportunities
Time: 2-3PM MST, Wednesday, April 14th
Download PDF Flyer: Seminar, April 14, 2021
Abstract: Join our interdisciplinary team for a presentation and panel/audience discussion of developing research on police officer decision-making under stress. Wearable technologies have the potential to monitor officers’ physiology and alert them to contexts where decision-making might be compromised. We discuss issues related to collecting continuous psychophysiological (e.g., heart rate) and vocal data; mapping physiology to stress and situational contexts; developing and deploying predictive algorithms and creating personalized machine-learning models; facilitating transmission, storage, and processing of intensive, high volume data; and integrating user feedback into all aspects of the process.
Dr. Nicole A. Roberts, School of Social & Behavioral Sciences
Dr. Nicholas Duran, School of Social & Behavioral Sciences
Dr. Yasin Silva, School of Mathematical & Natural Sciences
Dr. Ming Zhao, School of Computing, Informatics & Decision Systems Engineering
Machine Learning Day
You are invited to Arizona State University’s West campus for the Annual Machine Learning Day — an interactive virtual event!
Date held: Apr 09 2021 - 8:50am
Learn from cutting-edge researchers as they share innovative research on machine learning theory and methods in diverse domains, including applied statistics, biology, psychology, social science and ethics.
For graduate students: When you register you also have the option to present your research in a virtual poster session hosted through gather.town! The virtual conference will be held between 12:15–1:30.
Important things to know:
- Event is free.
- Registration is required
Download (PDF) flyer: Machine Learning Day Program flyer
Leveraging auxiliary information for detecting differentially expressed gene pathways
Speaker: Dr. Yue Wang, Assistant Professor of Biostatistics and Bioinformatics, School of Mathematical and Natural Sciences
When: March 3, 2021 02:00 PM Arizona
Download flyer: b2c2-wang.pdf
Abstract: Recent genetic studies have identified genes related to specific human diseases or traits. Besides marginal analysis of individual genes, analyzing biologically meaningful gene pathways, i.e., networks with nodes being genes and edges characterizing the presence/absence of the gene-gene interactions, may yield valuable insights. Identifying gene pathways that differ between conditions can be formulated as a multivariate hypothesis testing problem, but existing approaches handle the gene-gene correlations in inefficient ways, leading to inflated type I error rate and/or compromised power. In this paper, we propose a Hotelling's T2-type statistic, named the T2-DAG test, which efficiently leverages the edge information in the gene pathway through a linear structural equation model. We investigate asymptotic properties of the T2-DAG test under pertinent assumptions and compare the T2-DAG test with six existing methods under various simulation settings. We also apply the T2-DAG test to a lung cancer gene expression data set, and identity several interesting gene pathways that are relevant to different stages of lung cancer.
Speaker Bio: Dr. Wang is an Assistant Professor of Statistics in the School of Mathematical and Natural Sciences at Arizona State University. His research focuses on developing statistical methods for analyzing massive biological data to address cutting edge biological, clinical and public health related problems.
NSF National AI Institute for Student-AI Teaming
Speaker: Dr. Sidney D'Mello, Associate Professor in the Institute of Cognitive Science and Department of Computer Science at the University of Colorado Boulder
Time: 2:00 - 3:00 PM Arizona, February 17th (Wednesday), 2021
Download flyer: b2c2_dmello_flyer.pdf
Abstract: I will discuss our vision for the NSF National AI Institute for Student-AI Teaming. The Institute will: (1) develop the theories, Artificial Intelligence (AI) technologies, and know-how for creating next-generation collaborative learning environments composed of diverse students, teachers, and AI; (2) grow a diverse workforce of future AI researchers and practitioners by engaging 5,000 middle/high school students in innovative AI education through AI-enabled pedagogies; (3) serve as a national nexus point for empowering diverse stakeholders to engage in responsible co-design of student-AI collaborative technologies. This national institute brings together a geographically distributed team of researchers from nine Universities with partners from academia, K-12 school districts, and industry to address the central challenge of how to promote deep conceptual learning via rich socio-collaborative learning experiences for all students. To meet this challenge, the Institu te will reframe the role of AI in education, moving towards a future where AI is viewed as a social, collaborative partner that help students work and learn more effectively, engagingly, and equitably, while helping educators focus on what they do best: inspiring and teaching students. The Institute will adopt responsible innovation and polycultural approaches for developing ethical AI technologies by integrating foundational and use-inspired AI research across more than 12 interdisciplinary research areas spanning the computing, learning, cognitive and affective sciences. The Institute aims to lead the nation towards a future where all students routinely participate in rich and rewarding AI-enabled collaborative learning experiences that scale across a large number of classrooms, resulting in deeper student engagement and persistence in STEM, more inclusive classroom cultures, and significant improvements in learning outcomes.
Speaker Bio: Sidney D’Mello (PhD in Computer Science) is an Associate Professor in the Institute of Cognitive Science and Department of Computer Science at the University of Colorado Boulder. He is interested in the dynamic interplay between cognition and emotion while individuals and groups engage in complex real-world tasks. He applies insights gleaned from this basic research program to develop intelligent technologies that help people achieve to their fullest potential by coordinating what they think and feel with what they know and do. D’Mello has co-edited seven books and published more than 300 journal papers, book chapters, and conference proceedings (16 of which received awards at international conferences). His work has been funded by numerous grants and he currently serves as associate editor for Discourse Processes. D'Mello is the principal investigator of the AI Institute for Student-AI Teaming, which is one of the inaugural NSF National Artificial Intel ligence (AI) Research Institutes.
Pattern Recognition: Best Practices for (Interdisciplinary) Data Science
Speaker: Dr. Michael Simeone, Director of Data Science and Analytics for ASU Libraries
Time: 2:00 - 3:00 PM Arizona, November 18th (Wednesday), 2020
Download flyer: b2c2_simeone.pdf
Synopsis: In this presentation, Michael Simeone will cover best research practices when it comes to observing and interpreting patterns, from using data visualization to designing experiments and workflows.
Speaker Bio: Michael Simeone is a researcher interested in multidisciplinary data science. He currently serves as the director of Data Science and Analytics for ASU Libraries at Arizona State University. His research includes multidisciplinary data science, post-cybernetic culture and technology, analysis of human-technology networks, data visualization, and data-driven collaborations that bridge environmental sciences and humanities. Currently, he serves as a Domain Champion for Humanities for the Extreme Science and Engineering Discovery Environment. He received his doctorate in English from the University of Illinois at Urbana-Champaign.
Rescorla-Wagner Models with Dynamic Attention and Ensembles
Speaker: Dr. Joel Nishimura, School of Mathematical and Natural Sciences, ASU
Time: October 28 (Wednesday), 2020 2:00 - 3:00 PM Arizona
Download Flyer: b2c2-nishimura.pdf
Abstract: The Rescorla-Wagner (R-W) model is a discrete time stochastic process frequently used to model human learning, wherein an agent learns associations between cues and subsequent responses by dynamically updating cue associative strengths proportionally to a prediction error. We characterize the 'curse of dimensionality' for the R-W model and propose a method to overcome this limitation that uses dynamic attention to learn sparse signals. Given the difficulty in selecting features for prediction, dynamic attention faces challenges beyond those faced by the R-W model. Our new framework not only satisfies a constraint on the number of attended cues, it also performs better than the R-W model on a number of natural learning tasks, can correctly infers associative strengths, and focuses attention on predictive cues while ignoring uninformative cues.
Speaker Bio: Joel Nishimura is an Assistant Professor of Applied Mathematics at the West campus of ASU. He conducts research in several areas inside of and in the overlap between network science, dynamical systems and mathematical biology. A unifying feature of these systems is that simple rules can create complex behaviors and/or structures. He received his doctorate from Cornell University in 2013.
Whom we include and how they succeed
Presentation: “Whom we include and how they succeed”: A role for inclusion and equity in data-informed learning engineering
Speaker: Dr. Tom Fikes, Senior Researcher and Director of Research for the Action Lab at Arizona State University
Time: 2-3pm MST, October 14, 2020
Abstract: In this talk, I will begin with some definitions of “data science” and “learning engineering” – less for the purpose of arguing for the adoption of particular definitions, and more to begin a conversation about what the Action Lab is and does, with a goal of beginning a mutually beneficial research collaboration with B2C2 members around these topics in order to create even better learning environments for our ASU students (and particularly of our SSBS and New College students). I will argue that data science, learning engineering, and other similar useful professional identifications are structurally defined, but are always (if implicitly) embedded within functional contexts; and that social transformation requires a deep awareness on the part of data scientists, learning engineers, and allied agents and actors as to the nature of those functional contexts. Indeed, given the inevitable competition between functional contexts, we must go beyond merely being aware of them to cultivating and sustaining them. The second (and hopefully longer) portion of the talk will be around several data visualizations that highlight student learning, persistence, and retention at ASU. These visualizations will allow me to argue for a functional context that extends beyond “success for the average learner” (or for the institution) to one of equity, inclusion, and social justice. Not surprisingly, given the title, I will situate this within the context of the ASU charter and its language of inclusion and success.
Speaker Bio: Tom Fikes is Senior Researcher and Director of Research for the Action Lab at Arizona State University. He completed his Ph.D. at U.C. Santa Barbara where he studied cognitive psychology and cognitive neuroscience. Following postdoctoral work in mathematical modeling and cognitive science at Indiana University, he served teaching, research, and faculty administrative roles as professor of psychology and neuroscience at the University of Puget Sound in Tacoma, WA and Westmont College in Santa Barbara, CA. His current work centers on plying data science, modeling, visualization, and design thinking toward increased inclusivity, equity, and excellence in student learning.
Fun with Visualization in the Data Deluge
Speaker: Ross Maciejewski
September 16, 2020
1-2pm MST, Zoom
From smart phones to fitness trackers to sensor enabled buildings, data is currently being collected at an unprecedented rate. In this talk, we will explore how visualization can be leveraged to help us entertain fun and unique questions in the data deluge.By thinking about fun questions for datasets, we will demonstrate how visual computing can help build cross-domain collaborations, paving the way to discover new insights and challenges.
Ross Maciejewski is an Associate Professor at Arizona State University in the School of Computing, Informatics & Decision Systems Engineering and Director of the CAOE. His primary research interests are in the areas of geographical visualization and visual analytics. Professor Maciejewski is a recipient of an NSF CAREER Award (2014) and was named a Fulton Faculty Exemplar (2017) and Global Security Fellow at Arizona State. His work has been recognized through a variety of awards (2010, 2013, 2015, 2017, 2018). He currently serves as an Associate Editor for IEEE Transactions on Visualization and Computer Graphics.
Using Mathematical Models with Google Mobility Data to Predict COVID-19 in Arizona
September 2, 2020
1-2pm MST, Zoom
In June 2020, Arizona emerged as one of the country’s newest COVID-19 hot spots.In this talk, Dr. Wang will discuss a spatio-temporal forecasting model for COVID-19 cases with the help of human activity data from the Google Community Mobility Reports. The proposed model describes the combined effects of transboundary spread among county clusters and human activities on the spread of COVID-19. Haiyan Wang is a Professor of Applied Mathematics at the School of Mathematical and Natural Sciences at Arizona State University. He obtained his doctorate in mathematics and master in computer science at Michigan State University in 1997. Before coming to the West campus of ASU in 2005, Dr. Wang had worked in industry as a software engineer for many years. Dr. Wang’s research interests include applied mathematics, differential equations, mathematical biology and data science.