DINS Seminar Series: 2022-23

Welcome to the DINS Seminar Series for AY 2022-23. This page will be updated as more information about specific talks becomes available.

The Department of Informatics and Networked Systems (DINS) is pleased to invite you to join us for a stimulating series of lecture events, the DINS Seminar Series. Taking place regularly throughout the 2022-2023 Academic Year, the series will feature a variety of experts from academia and industry who will discuss their leading-edge work on topics at the intersections of information and data, organizations and society, and systems and networks. The breadth of topics covered in the series reflects the diversity of research and educational opportunities within the Department. 

Quantifying Political Polarization on a Network Using Generalized Euclidean Distance

Speaker: Michele Coscia, IT University of Copenhagen

October 19, 2022
11.00 a.m. - 12.00 noon
Online (Join Zoom Meeting: https://pitt.zoom.us/j/93599179854  Meeting ID: 935 9917 9854)

Abstract:  An intensely debated topic these days is whether political polarization on social media is on the rise or not. This question can only be investigated if we have a measure for polarization. Such a measure should take into account how extreme the opinions of the people are, how much they organize into echo chambers, and how these echo chambers organize in the network. The most popular ways of estimating polarization are insensitive to at least one of these factors, thus they cannot conclusively clarify the opening question. We propose a measure of ideological polarization which can capture these factors. The measure is based on the Generalized Euclidean (GE) distance, which estimates the distance between two vectors on a network, e.g., vectors representing people’s opinion. This measure can fill the methodological gap left by the current state of the art, and leads to useful insights when applied to real-world debates happening on social media and to data from the US Congress.

Gender disparities and the Glass Ceiling effect in science

Speaker: Kristina Lerman, Principal Scientist at University of Southern California Information Sciences Institute

October 26, 2022
11.00 a.m. - 12.00 noon
Online (https://pitt.zoom.us/j/94529250331)

Meeting ID: 945 2925 0331

Abstract: Gender disparities persist in science, systematically reducing career opportunities for women. As a result, women remain a small minority in many fields, especially in senior positions. The dearth of elite women scientists, in turn, leaves fewer women to serve as mentors and role models for the younger generation. We explore gender disparities in citations, showing that women receive less recognition for their work relative to men, and that this cannot be explained purely by their minority status. Instead, gender disparity arises from biased individual preferences about who to cite, that are amplified by cumulative advantage. We present a model for the growth of citations that captures this mechanism and analyze it to show that its predictions align with real-world observations. In the second part of the talk, I present a study of prominent scholars who were elected to the National Academy of Sciences. I identify gender disparities in the structure of citation networks of these researchers and show that these differences are strong enough to accurately predict the scholar's gender. These results provide further evidence that a scholar's gender plays a role in the mechanisms of success in science.

Resilience of Urban Sociotechnical Systems to Disasters, Pandemics, and Societal Changes 

Takahiro Yabe

Speaker: Takahiro Yabe, Postdoctoral Associate in the Connection Science group, SSRC, MIT

November 2, 2022
11:00 am to 12 noon
Online: https://pitt.zoom.us/j/94231244662

Meeting ID: 942 3124 4662

Abstract: Cities are the central engines of productivity, innovation, and cultural diversity, owing to their ability to foster dense social and economic connections among people and organizations. However, cities are also at the forefront of unprecedented challenges, including increased frequency of climate change induced disasters, aging infrastructure systems, and growing inequality and segregation. To build urban resilience to such challenges, we need to better understand the cascading socioeconomic impacts of shocks, which are undergirded by complex interdependencies between social networks, urban environments, and online systems. Leveraging the increasing availability of large-scale human behavior data collected from mobile devices (e.g., mobile phone GPS, social media, web search) I study the resilience of urban sociotechnical systems using a data-driven complex systems dynamics approach. In this talk, I will particularly focus on research on 1) resilience of cities to disasters, focusing on the impacts of complex interdependencies between social dynamics and infrastructure systems (based on papers in PNAS and Sustainable Cities and Society), and 2) resilience of the diversity of urban social contact networks during the pandemic (arXiv preprint available). 

Bio: Taka is a Postdoctoral Associate at the MIT Institute for Data, Systems, and Society (IDSS) and Media Lab mentored by Alex 'Sandy' Pentland and Esteban Moro. Taka’s research lies in the intersections of civil engineering, computational social science, and urban science. His research develops data-driven urban sociotechnical system models for analyzing large-scale human behavior data, to better understand collective social dynamics during disruptions, and to improve the resilience of communities and cities to urban shocks. He received his Ph.D. from Purdue University, advised by Satish V. Ukkusuri, and his Masters and Bachelor Degrees from the University of Tokyo. Taka is also passionate in transforming research to policy, and previously was a Data Science Consultant for the World Bank, working on human mobility data analytics for urban disaster risk management and transportation resilience modeling. 


Creating Critical Technologists: Shifting Computing Education towards JusticeAngela Stewart U of Pitt

Speaker: Angela Stewart, Assistant Professor, Department of Informatics and Networked Systems, School of Computing and Information, University of Pittsburgh

November 9, 2022
11:00 am to 12 noon
Meeting ID: 971 2860 8499

Abstract: Computing education largely focuses on equipping learners with technical competencies to be part of the future workforce. However, this approach has led to technologists who lack the skills to critically reflect on the ways technology might uphold systems of oppression. In this talk, I will discuss a reimagined vision for justice-oriented computing education. I will present findings from several design-based research studies in after-school computing education for girls. I will discuss how principles from these studies can make for more equitable computing education.

Bio: Angela joined the University of Pittsburgh in 2022, with a joint appointment in the School of Computing and Information and Learning Research and Development Center. Prior to that, she was a Postdoctoral Fellow in the Human-Computer Interaction Institute at Carnegie Mellon University. She received a PhD in Computer Science from the University of Colorado Boulder (2020) and a Bachelor of Software Engineering from Auburn University (2015). Angela conducts research at the intersection of the learning sciences, artificial intelligence, and human-computer interaction. She uses multimodal data to understand students' social and cognitive states, particularly in collaborative STEM learning. She also creates equitable educational spaces by designing technologies that support the agency of students and teachers. Angela applies a culturally-responsive lens to her research, with a particular focus in emboldening Black girls' design of transformative technologies. Angela was named a 2021 - 2022 Emerging Scholar by the International Society of the Learning Sciences.


The Lived Experiences Measured Using Rings Study (LEMURS)

Speaker: Christopher M. Danforth, Professor, Department of Mathematics & Statistics, College of Engineering and Mathematical Sciences,

University of Vermont

November 16, 2022
11:00 am to 12 noon
Virtual -- Join Zoom Meeting   https://pitt.zoom.us/j/92863524224

Meeting ID: 928 6352 4224 

Abstract: Building on a decade of work quantifying mood using social media activity, this talk will describe our group’s new longitudinal wearables study of health and well-being. A cohort of 600 first-year students at the University of Vermont have been recruited to take part in an experiment incentivizing behaviors associated with human flourishing. During the Spring semester of 2023, three groups of 150 students will be randomly assigned to engage in expert guided (a) exercise classes, (b) nature experiences, and (c) group talk therapy. Changes in stress, mental health, and sleep will be assessed through a series of weekly surveys deployed through a dedicated app, as well as continuous physiological heart-rate monitoring through the Oura Ring, and compared to a fourth control group of 150 students.

Bio: Professor Chris Danforth received a B.S. in Mathematics & Physics from Bates College in 2001, and a Ph.D. in Applied Mathematics & Scientific Computation from the University of Maryland, College Park in 2006. His early work applied Chaos Theory to improve forecasts made by numerical weather models, and his current work focuses on sociotechnical systems. He is the co-inventor of http://hedonometer.org, an instrument measuring daily happiness based on social media, and has also developed algorithms to identify predictors of depression from Instagram photos. Along with Peter Sheridan Dodds, Danforth runs the Computational Story Lab research group and the Vermont Complex Systems Center. Danforth is Director of the Vermont Advanced Computing Center, and his work has also been funded by NIH, NASA, NOAA, DARPA, DOE, and the MITRE Corporation. Danforth has co-authored over 100 peer-reviewed publications applying mathematical techniques to many fields including atmospheric science, linguistics, psychology, literature, finance, physics, engineering, and biochemistry. He has advised over 50 research dissertations including 20 PhD students, 20 MS students, and 15 undergraduate thesis students. Descriptions of his projects are available at his website: http://cdanfort.w3.uvm.edu


Socially and Ethically Responsible AI for Sustainable Development: Bringing Invisible Millions at the Center of the AI Revolution

Speaker: Neil Gaikwad, PhD Candidate, MIT

December 14, 2022
11:00 am to 12 noon

Virtual -- Join Zoom Meeting: https://pitt.zoom.us/j/99308939022

Abstract: How should we design AI/ML technologies to benefit the world's poorest m who are invisible in mainstream datasets and not only experiencing the disproportionate impact of climate change and structural inequalities but also algorithmic harms? AI/ML models, trained and evaluated with highly curated datasets and standard benchmarks, demonstrate remarkable computational efficiency in lab settings or online human-subject studies but are ineffective when deployed in the real world. Designing Ethical AI/ML systems for scientifically informing high-stake policy decisions is unquestionably one of the most difficult challenges. In this talk, I present a research program focused on Socially and Ethically Responsible AI for Sustainable Development. By bringing digitally invisible at-risk communities to the center of human-AI collaboration, the scholarship ensures fairness, accountability, transparency, and ethics as an intrinsic part of human-centered AI rather than afterthought optimization. With real-world deployments, I demonstrate what Responsible AI looks like from the perspective of the most vulnerable (e.g., smallholder farmers, racial minorities, gig workers, etc.). This publicly engaged research program shifts the paradigm of the AI revolution to make a newly realized sociotechnical world more inclusive and sustainable.

Bio: Neil Gaikwad is a Ph.D. candidate at MIT, where he is a Human Rights and Technology Fellow and Social and Ethical Responsibilities of Computing Scholar. His research in computational sustainability straddles the interface of Ethical AI and Policy for promoting global inclusion, equity, and societal development. This work has been published in premier artificial intelligence and human-computer interaction conferences (AAAI, KDD, CHI, CSCW, UIST), journals (PNAS), and featured in venues such as The New York Times, New Scientist, Bloomberg, WIRED, and Wall Street Journal. Neil has been recognized with numerous awards in science and engineering, including Facebook Ph.D. Fellowship, MIT Graduate Teaching Award, and Rising Star in Data Science by the University of Chicago, and the Karl Taylor Compton Prize, MIT’s highest student award. He has mentored more than 20 students who published impactful papers, won prestigious fellowship awards, pursued careers in research, and shifted the discourse on AI fairness and racial equity. Neil holds a master’s degree from the School of Computer Science at Carnegie Mellon University. Before MIT, he was a data scientist on Wall Street. 

Science of science, law of law, and patterns of patents: universal citation dynamics in knowledge systems

Speaker: Yong-Yeol (YY) Ahn, Associate Professor, Center for Complex Networks and Systems Research, Luddy School of Informatics, Computing, and Engineering, Indiana University Bloomington

January 18, 2023

11:00 am to 12 noon

In-person -- Room 316 IS Building, 135 North Bellefield Avenue, Pittsburgh, PA 15213

Zoom option: Join Zoom Meeting: https://pitt.zoom.us/j/93743886499
Meeting ID: 937 4388 6499


Citation is a fundamental way for humans to acquire and expand on existing knowledge. Although many laws and regularities of citation dynamics have been discovered from scientific citations, it is unclear whether and to what extent these regularities are inherent in how humans seek, use, and create knowledge. We show that, despite many stark differences between these systems, the citation dynamics in science, law, and patents share universal patterns. Given the differences in procedure and incentives that exist between judges, inventors, and scientists, our findings suggest that universal citation dynamics may be innate to any cumulative human knowledge system. Our model demonstrates that the evolution of collective attention and a handful of fundamental mechanisms can produce observed universal patterns of citation dynamics.


Designing a Critical Data Literacy Toolkit: Centering Youth Experiences

Speaker: Aditi Mallavarapu,Technology and Learning Sciences Postdoctoral Researcher, Digital Promise; Visiting Scholar, the Learning Research and Development Center (LRDC) at the University of Pittsburgh

February 1, 2023
11:00 am to 12 noon
In person -- Room 538-539 130 North Bellefield Building (Conference Room)

Zoom information for those who wish to attend remotely: Join Zoom Meeting: https://pitt.zoom.us/j/98850768917

Abstract: Data literacy (DL) competencies can play a critical role in addressing the underrepresentation of Black youth in technology workforce pathways. Designing inclusive data literacy resources and technology to match the expectations and needs of historically marginalized youths is essential for reimagining a more inclusive technological workforce. In this talk, I discuss our partnership with nine Black high school students from a historically marginalized community to reimagine the design of a critical DL toolkit, positioning them as advisors for our research project, DATA (Data Analysis to Action) project. Through youth participatory action research (YPAR) the youth engaged in facilitated data advocacy activities analyzing data via a data analysis technology and creating artifacts for advocacy using community-based data. Our advisors identified refinements for DL technology design and activities that better align with their needs. Further centering critical DL, we analyzed advisors’ interactions during the hands-on activities through the data feminism lens to understand how the activities manifest critical thinking with data. We identified technical affordances that can support the advisors in this endeavor. Our major findings highlight that: engaging youth with data that aligns with their lived experiences can empower the youth; moreover, flexible technology and curricular design that supports critical thinking with data can expose the youth to the potential of using data and technology as an empowering tool for advocacy. I discuss these findings and the implications about the data, learning technology and curricular activities that the youth highlighted.

Bio: Aditi Mallavarapu is a Technology and Learning Sciences Postdoctoral researcher at Digital Promise, working as a part of the Center for Integrated Research in Computing and Learning Sciences (CIRCLS) where she specializes in creating data-driven resources to drive research innovation for researchers in the community. She is also a visiting scholar at the Learning Research and Development Center (LRDC) at the University of Pittsburgh, where she uses participatory design to develop culturally responsive technology for young learners to learn data literacy and advocacy skills. She earned her PhD in Computer Science from University of Illinois at Chicago. Her work uniquely investigates the underexplored space of applying human-centered learning analytics and machine learning techniques for exploration-based learning that takes place in complex open-ended learning environments (e.g., museum exhibits). Combining the human-centered approach working alongside interdisciplinary researchers, educators, learners, and policymakers, with the technical data-driven approaches, Aditi reimagines open-ended learning environments as data-driven systems designed to engage learners with real-world complex problems.

The Moral Machine Experiment

Speaker: Edmond Awad, Senior Lecturer in the Department of Economics and the Institute for Data Science and Artificial Intelligence, University of Exeter

February 8, 2023
11:00 am to 12 noon

Topic: DINS Seminar Series -- Edmond Awad
Time: Feb 8, 2023 11:00 AM Eastern Time (US and Canada)

Join Zoom Meeting

Abstract: I describe the Moral Machine, an internet-based serious game exploring the many-dimensional ethical dilemmas faced by autonomous vehicles. The game enabled us to gather 40 million decisions from 3 million people in 200 countries/territories. I report the various preferences estimated from this data, and document interpersonal differences in the strength of these preferences. I also report cross-cultural ethical variation and uncover major clusters of countries exhibiting substantial differences along key moral preferences. These differences correlate with modern institutions, but also with deep cultural traits. I discuss how these three layers of preferences can help progress toward global, harmonious, and socially acceptable principles for machine ethics. Finally, I describe other follow up work that build on this project.

Bio: Edmond Awad is a Senior Lecturer (Assistant Professor) in the Department of Economics and the Institute for Data Science and Artificial Intelligence at the University of Exeter. He is also an Associate Research Scientist at the Max Planck Institute for Human Development, a Turing Fellow at the Alan Turing Institute, and a Founding Editorial Board member of the AI and Ethics Journal (Springer Nature). Before joining the University of Exeter, Edmond was a Postdoctoral Associate at MIT Media Lab (2017-2019). In 2016, Edmond led the design and development of Moral Machine, a website that gathers human decisions on moral dilemmas faced by driverless cars. The website has been visited by over 9 million users, who contributed their judgements on 100 million dilemmas. Another website that Edmond co-created, called MyGoodness, collected judgements over 3 million charity dilemmas. Edmond’s work appeared in major academic journals, including Nature, PNAS, and Nature Human Behaviour, and it has been covered in major media outlets including The Associated Press, The New York Times, The Washington Post, Der Spiegel, Le Monde and El Pais. Edmond has a bachelor degree (2007) in Informatics Engineering from Tishreen University (Syria), a master’s degree (2011) in Computing and Information Science and a PhD (2015) in Interdisciplinary Engineering from Masdar Institute (now Khalifa University; UAE), and a master’s degree (2017) in Media Arts and Sciences from MIT. Edmond’s research interests are in the areas of AI, Ethics, Computational Social Science and Multi-agent Systems.

Designing Human-centered Tools to Navigate the New World of Work

Speaker: Isabella Loaiza, PhD student and Research Assistant, Human Dynamics Group, MIT Media Lab

February 15, 2023
11:00 am to 12 noon
Virtual --  https://pitt.zoom.us/j/96042627526


Technological progress and the rise of remote, hybrid, and gig work are reshaping our traditional notions of jobs and careers. As the new world of work offers increased flexibility and a growing number of potential paths, it is becoming more difficult for humans to make informed career decisions. In this talk, I will explain how I use Value-Sensitive Design (VSD) and large-scale occupation data to build two tools to help individuals navigate the new AI-driven labor markets. The first tool addresses economic integration for migrant populations, and the second focuses on US workers looking to relocate to advance their careers.

Isabella is a Ph.D. student in the Human Dynamics Group at the MIT Media Lab. Her work focuses on understanding how to design tools to help people work with, instead of against, machines while reinforcing agency and self-determination. Her main focus areas are migrant and refugee relocation, urban mobility, and career design.

Natural Language Processing (NLP) with User Feedback

Speaker: Xiaozhong Liu, School of Informatics, Computing and Engineering at Indiana University Bloomington

March 1, 2023

11 am to 12 noon

Virtual -- Join Zoom Meeting: https://pitt.zoom.us/j/93422077271

Meeting ID: 934 2207 7271

In an active cyberspace ecosystem, Natural Language Processing (NLP) algorithms and tasks are essential components. The massive text content is often accompanied with abundant heterogeneous data, e.g., click logs, knowledge graphs, chronological search history, and computational user profiles, which provide us opportunities of taking advantage of data heterogeneity for further advancing NLP algorithms. By leveraging various kinds of user-generated data and sophisticated deep learning methodologies, we proposed a number of algorithms and applications to address different novel and classical research questions, and, moreover, we also successfully validated and adopted them on industry platforms to serve millions of users. In this talk, I will present three cases: 1) Cybercrime Detection with Berrypicking Behaviors, 2) Natural Language Generation (NLG) with Click-logs, and 3) Multi-role Dialogue Mining with Multi-task Pretraining. In these works, user-centric heterogeneous data and text information are jointly encapsulated by using innovative representation learning and multi-task learning.

Bio: Xiaozhong Liu is an Associate Professor from School of Informatics, Computing and Engineering at Indiana University Bloomington. He also serves as senior consultant to Alibaba DAMO Academy, the research arm of Alibaba, and leads large-scale NLP projects. He has published more than 150 papers in leading computer science conferences and information science journals, e.g, PNAS, JASIST, AAAI, SIGIR, IJCAI, EMNLP, ACL and WWW, and holds nine patents in AI and NLP. His areas of research interest include Data Science, NLP, Explainable AI, Graph Mining, Cybercrime and Security, and Computational Social Science. Currently, his algorithm APIs are called more than 170 million times per day on different active eCommerce and Web Search platforms and effectively outperforms predecessor methods by 10% on average.

A Global Perspective of Scientific Publications

Speaker: Bedoor Alshebli, Assistant Professor, Computational Social Science, NYU Abu Dhabi

March 15, 2023
11:00 am to 12 noon
In person -- Room 538-539 Conference Rooms in 130 North Bellefield Avenue (across the street from the IS Building). Refreshments will be served.

Virtual: Join Zoom Meeting https://pitt.zoom.us/j/92448706592

Meeting ID: 924 4870 6592

This talk presents an overall look at Bedoor AlShebli's research to date, where she examines scientific publications from a global perspective. The first part of the talk analyzes six publishers---Frontiers, Hindawi, IEEE, MDPI, PLOS, and PNAS---to understand racial and geographical disparities in terms of: (i) editorial board composition; (ii) the time spent between the submission and acceptance of a manuscript; and (iii) the number of citations a paper receives relatively to textually-similar papers. The second part of the talk examines gender inequality in the editorial boards of Elsevier journals across disciplines over the past decades. The third part focuses on homophily in scientific collaborations and examines the relationship between the ethnic diversity of co-authors and the impact of the paper they produce. The fourth and final part focuses on AI research across the globe, examining the inter-city networks of citations, collaborations, and scientists’ migrations to uncover dependencies between Eastern and Western cities worldwide.

Bio: Bedoor AlShebli is an Assistant Professor of Computational Social Science at New York University Abu Dhabi, UAE. She received her PhD in Interdisciplinary Engineering in 2017 from Masdar Institute of Science and Technology, and her MSc in Computer Science from the University of Illinois at Urbana-Champaign. Bedoor's research focuses on using data science techniques to study social phenomena, with a particular emphasis in the Science of Science. She is interested in the social and economic benefits of diversity, as well as the dynamics of social interaction and cohesion, and frames social science problems in the contexts of data science, big data, and applied machine learning. Her work appeared in major academic journals, including Nature Human Behavior, Nature Communications, Science Advances, and the Proceedings of the National Academy of Sciences. Cumulatively, her work contributes to the fields of computational social science, data science, and machine learning.

Speaker: Peter Sheridan Dodds, Professor, Department of Computer Science, Director of Vermont Complex Systems Center, University of Vermont 


Pathway of Innovation

Speaker: Hyejin Youn, Associate Professor, Management & Organizations, Kellogg School of Management, Northwestern University

March 29, 2023
11:00 am -- 12 noon
In person -- Room 538-539 Conference Rooms in 130 North Bellefield Avenue (across the street from the IS Building). Refreshments will be served.

To attend virtually, Join Zoom Meeting https://pitt.zoom.us/j/93680987565

 Meeting ID: 936 8098 7565

Abstract: Is innovation dominated by individual actions or should it be understood as a collective process to make the whole greater than the sum of its parts? Innovation that brings about new knowledge is often described as individual search processes, exploiting and exploring a complex landscape of concepts, ideas, and items to find their novel connections. Then, where does the complex landscape come from? As we search the space individually, the landscape is collectively structured, maintained by, and shared with the society, which further down to open up new adjacent possible or close down existing possibilities to individuals' future invention activities. Such co-evolution of constraints explains the full of contingent and idiosyncratic cases, resulting in the unpredictable nature of innovation. Innovation is path dependent. However, what if past-paths are well-understood? Does this mean we can predict the future innovation from the past innovation? It is indeed the case that there has been evidences of predictive innovations that contradict the very premises of innovation. Here we construct a stochastic model of co-evolution between individual actions and the landscape. The landscape is an accumulation of our collective past paths. Our model shows individual activities who have incentive to be aligned often create a convention toward a paradigm, becoming an increasing inertia moving toward a particular dynamic direction at a larger scale than that of individual dynamics. As a result, the landscape becomes more modular and rugged structure. By constructing an innovation phase diagram of counter-factual spaces, and by analyzing two independent empirical datasets – almost two centuries of the U.S. patents and four decades of scientific publications – we demonstrate that both science and technology evolve at the edge between the exploitation strategy and exploration strategy. We show the individual innovative process shapes the underlying knowledge landscape, but also their future actions are shaped by the very landscape they have been shaping. Such co-evolution processes between individuals and the collective structure seem to leave and pave the path-dependent trajectory of innovation, which explains predictable innovation.

Bio: Dr. Youn is an Associate Professor, Management & Organizations, Kellogg School of Management at Northwestern University, and Northwestern Institute on Complex Systems (NICO). She is also an external faculty at Santa Fe Institute. Her research aims to develop a mathematical and computational framework to understand complex systems. 



Sarah Bana, Assistant Professor of Management Science, Chapman University, and Digital Fellow, Stanford Digital Economy Lab

April 5, 2023, 11:00 am to 12 noon

in-person -- Room 538-539 Conference Room, 130 North Bellefield Avenue (across from the IS Building). Refreshments will be served. 

To attend virtually: Topic: DINS Seminar Series -- Sarah Bana

Time: Apr 5, 2023 11:00 AM Eastern Time (US and Canada)

Join Zoom Meeting: https://pitt.zoom.us/j/94544266139

Meeting ID: 945 4426 6139

Abstract: How valuable are various IT skills? How do firms adjust their workforce composition after adopting new technology?  Addressing these questions and others on the role of technology in the labor market has been hindered by data limitations. This project explores an innovative approach by transforming the text content of job postings into vector representations, referred to as work2vec. 

In the first paper, I apply natural language processing (NLP) techniques to build a model that predicts salaries from the job posting text. This follows the rich tradition in the economics literature of estimating wage premia for various job characteristics by applying hedonic regression. Using an attribution method called integrated gradients, I find that Cloud Computing and System Design and Implementation skills are amongst the most valued IT skills in 2019, while Database Administration and Data Management have more moderate effects. 

In the second paper, my co-authors Erik Brynjolfsson, Daniel Rock and Sebastian Steffen and I develop a spatial representation of work using a variational autoencoder (VAE). We use this representation to measure how occupations and the job space as a whole are evolving over time, along with the specific question of machine learning (ML) adoption. We find that firms adopting ML tools significantly alter the composition of job postings towards occupations that appear more complementary to ML tools. 


Bio: Sarah Bana is an Assistant Professor of Management Science at Chapman University and a Digital Fellow at the Stanford Digital Economy Lab. Dr. Bana's research focuses on the effect of technology on the labor market. This involves combining novel sources of text data, like job postings and syllabi, with natural language processing methods to develop innovative insights. Her research has been published in outlets such as the Journal of Econometrics, the Journal of Policy Analysis and Management, and Sloan Management Review, and featured by the BBC Business Daily. Dr. Bana's research has been funded by the Russell Sage Foundation. She received her Ph.D. in economics at the University of California, Santa Barbara. Her dissertation received Honorable Mention for the Upjohn Institute’s best Ph.D. dissertation on employment-related issues.


Engineering Secure and Privacy-Preserving Systems

Jaideep Vaidya, Distinguished Professor of Computer Information Systems at Rutgers University and the Director of the Rutgers Institute for Data Science, Learning, and Applications

April 12, 2023

11:00 am - 12:00 noon

in-person -- 538-539 Conference Room in 130 North Bellefield Avenue, across the street from the IS Building. Refreshments will be served.

To attend virtually -- https://pitt.zoom.us/j/93846645007

Meeting ID: 938 4664 5007

Abstract: In the current digital age, data is continually being collected by organizations and governments alike. While the goal is to use this data to derive insight and improve services, the ubiquitous collection and analysis of data creates a threat to privacy. Furthermore, the digitization and centralization of data creates attractive targets for cyber criminals, with security breaches harming both individuals and organizations. In this talk, we discuss the current state of the art in engineering secure and privacy-preserving systems, presenting our recent work on access control configuration and management, synthetic data generation, and privacy-preserving analytics which is crucial to maintaining organizational security and privacy.

Biography: Jaideep Vaidya is a Distinguished Professor of Computer Information Systems at Rutgers University and the Director of the Rutgers Institute for Data Science, Learning, and Applications. He received the B.E. degree in Computer Engineering from the University of Mumbai, the M.S. and Ph.D. degree in Computer Science from Purdue University. His general area of research is in security, privacy, data mining, and data management. He has published over 200 technical papers in peer-reviewed journals and conference proceedings, and has received several best paper awards from the premier conferences in data mining, databases, digital government, security, and informatics. He is an ACM Distinguished Scientist, an IEEE and AAAS Fellow and served as the Editor in Chief of the IEEE Transactions on Dependable and Secure Computing.

Towards Inclusive and Equitable Language Technologies

Malihe Alikhani, Assistant Professor, Computer Science, School of Computing and Information, University of Pittsburgh

April 19, 2023

11 am 12 noon

in-person -- 538-539 Conference Room in 130 North Bellefield, across from the IS Building. Refreshments will be served.

To attend virtually: Join Zoom Meeting --  https://pitt.zoom.us/j/93909536513

Meeting ID: 939 0953 6513

Abstract: The societal needs of language technology users are rapidly changing alongside the technological landscape, calling for a deeper understanding of the impact of natural language processing models on user behaviors. As these technologies are increasingly deployed to users, it is ever more important to design culturally responsible and inclusive language technologies that can benefit a diverse population.

Toward this goal, I present two directions: 1) Discourse-aware models for more inclusive and productive social media moderation methods, and 2) Equitable machine learning frameworks for multimodal communication. Finally, I describe my research vision: Building inclusive and collaborative communicative systems by leveraging the cognitive science of language use alongside formal methods of machine learning.

Bio: Malihe Alikhani is an Assistant Professor of computer science in the School of Computing and Information at the University of Pittsburgh. She earned her Ph.D. in computer science with a graduate certificate in cognitive science from Rutgers University in 2020. Her research interests center on using representations of communicative structure, machine learning, and cognitive science to design practical and inclusive NLP systems for social good. Her work has received multiple best paper awards at ACL 2021, UAI2022, INLG2021, and UMAP2022 and has been supported by DARPA, NIH, Google, and Amazon.

Overview of Learning Quantum States

Srinivasan Arunachalam, Senior Research Scientist, IBM Almaden

Wednesday, April 26, 2023

11 am to 12 noon

Virtual: Join Zoom Meeting, Meeting ID: 959 7218 6621 https://pitt.zoom.us/j/95972186621

Abstract: Learning an unknown n-qubit quantum state is a fundamental challenge in quantum computing theory and practice. Information-theoretically, it is well-known that tomography requires exponential in n many copies of an unknown state in order to estimate it upto small trace distance. But a natural question is, are there models of learning where fewer copies suffice and are there interesting classes of states that can be learned with fewer copies? In this talk I will discuss the following results: (1) learning local Hamiltonians on n qubits using poly(n) many samples of the quantum Gibbs state, (2) In the past few years, there have been various learning models introduced to capture the learnability of quantum states; here I will overview many recent results and discuss various equivalences between these learning models. Both these works pave the way towards a more rigorous application of using machine learning techniques to learning quantum states.

Bio:Srinivasan Arunachalam is a Senior Research Scientist at IBM Almaden. Prior to this, he was a Postdoctoral Researcher at MIT. He received his Ph.D. in 2018 from QuSoft, Netherlands and his Master's degree from University of Waterloo. His research interests are in quantum learning theory, algorithms and complexity theory. 

Efficient Evaluation of Attribute-Based Access Control Policies

Shamik Sural, Department of Computer Science and Engineering, Indian Institute of Technology (IIT) Kharagpur

Friday, April 28, 2023

3:00 pm to 4:00 pm

In-Person:  Room 538-539 Conference Room, 130 N. Bellefield Avenue (across the street from the IS Building). No virtual option available. 

Abstract: Access control mechanisms are used by organizations to mitigate the risk of unauthorized access to data, resources and systems. For traditional information systems that deal only with a pre-specified set of users, access control models like Discretionary Access Control (DAC), Mandatory Access Control (MAC) and Role-Based Access Control (RBAC) work satisfactorily. The primary limitation of these traditional models is their significant dependence on user identity for making access decisions. Owing to this, such models are found to be unsuitable for dynamic situations, where unknown users from various domains may have to be given access. Further, an inherent lack of extendibility makes it difficult to consider the context in which the access request is made. To handle these requirements, the Attribute-Based Access Control (ABAC) model has recently been proposed.

In ABAC, a user is permitted or denied access to an object based on a set of rules (together called an ABAC Policy) specified in terms of the values of attributes of the different types of entities, namely, user, object and environment. Efficient evaluation of these rules is therefore essential in ensuring decision making at on-line speed when an access request comes. Sequentially evaluating all the rules in a policy is inherently time consuming and does not scale well with the size of the ABAC system and the frequency of access requests. This problem, which is quite pertinent for practical deployment of ABAC, has so far received little attention from the research community.

In this talk, we introduce two variants of a tree data structure for representing ABAC policies, which we name as PolTree. In the binary version (B-PolTree), at each node of the tree, a decision is taken based on whether a particular attribute-value pair is satisfied or not. The n-ary flavor (N-PolTree), on the other hand, grows as many branches out of a given node as the total number of possible values for the attribute being checked at that node. Extensive experimental evaluation with diverse data sets shows the scalability and effectiveness of this approach.

Bio: Shamik Sural is a full professor in the Department of Computer Science and Engineering, Indian Institute of Technology (IIT) Kharagpur. He received the Ph.D. degree from Jadavpur University, Kolkata, India in the year 2000. Before joining IIT in 2002, Shamik spent more than a decade in the Information Technology industry working in India as well as in Michigan, USA.
Shamik was a recipient of the Alexander von Humboldt Fellowship for Experienced Researchers in 2009, which enabled him to carry out research at TU Munich, Germany. He spent the Fall 2019 semester at Rutgers University as a Fulbright scholar engaged in both teaching and research. He is also an ACM Distinguished Speaker. Shamik is a senior member of IEEE and has previously served as the Chairman of the IEEE Kharagpur section. He is currently serving on the editorial boards of IEEE Transactions on Dependable & Secure Computing and IEEE Transactions on Services Computing. His research interests include computer security and data science.