Fall Term 2023

The colloquia are held in English and take place from 5.15 to 6.30 p.m. in room 2.A.01 at the Department of Informatics (IfI), Binzmühlestrasse 14, 8050 Zürich. 

Visiting a colloquium is free of charge and does not require registration. 

Details about the format of the talk shall be checked always just ahead of a certain presentation date.

If you have further questions please contact Karin Sigg.
 

Flyer to download (PDF, 231 KB)

The IFI colloquium talks will be held at the date indicated below. They start at 5:15 p.m. or see time below date.

* If you would like to get access to the talk please send an email to  studies@ifi.uzh.ch

21.09.2023 – Empowering Users and Communities in Social Media Governance

Speaker: Prof. Amy Zhang, Ph.D.

Host: Dr. Cristina Sarasua / Prof. Abraham Bernstein, Ph.D.

Abstract

Content moderation on online social platforms is broken. As a result of their inability to moderate adequately, platforms have become host to all manner of undesirable content and behavior, from hate and harassment targeted at marginalized groups to the spread of misinformation that is harmful to society. In this talk, I argue that many of today’s content moderation systems fall short because they are operating from two flawed premises—that it is possible to come up with a single set of policies to enforce consistently across a large and diverse user base, and that users should have frictionless power to post and yet no responsibility towards (or power to shape) the collective governance of a social space. Instead, my research focuses on how to design the governance structures, processes, and tooling of online social platforms so that they empower users and communities to have greater agency and voice in governance. My contributions include: 1) new tools at the community level to support more powerful and flexible collective design and execution of policies, 2) new tools at the user level to support individuals with moderating what they personally see, and 3) reimagining processes at the platform level to legitimize platform-wide decisions.

Bio

Amy X. Zhang is an assistant professor at University of Washington's Allen School of Computer Science and Engineering, where she leads the Social Futures Lab, a group dedicated to reimagining social and collaborative systems to empower people and improve society. Her work has received awards at ACM CSCW and ACM CHI, and she has been a Google Research Scholar, a Belfer Fellow at the ADL, a Berkman Klein Fellow, a Google PhD Fellow, and an NSF Graduate Research Fellow. Besides her work at UW, she is also currently a research consultant at AI2 with Semantic Scholar, and prior to UW, she was a Stanford postdoctoral researcher after completing a PhD at MIT CSAIL, where she received the George Sprowls Best Thesis Award at MIT in computer science. She received an MPhil in Computer Science at the University of Cambridge on a Gates Fellowship and a BS in Computer Science at Rutgers University, where she was captain of the Division I Women's tennis team.

28.09.2023 – Use-Inspired Research

Speaker:  Prof. David C. Shepherd, Ph.D.

Host: Prof. Dr. Thomas Fritz

Abstract

“The Pasteur’s quadrant approach implies that it is possible to couple, simultaneously, the development of fundamental science to the solution of problems important to society."
-Donald Stokes

Science is often artificially segmented into curiosity-driven research and completely applied research. However, there is a hybrid type of research, often called use-inspired research, that starts from important societal problems, but invents new fields, techniques, and fundamental knowledge to attack these problems. During my career, and especially while in academia, this is how I have conducted my research. In this talk, I will explain the basics of use-inspired research, why it is important to our scientific landscape today, and illustrate it with my research projects spanning robotics, CS education and outreach, and healthcare.  

Bio

David Shepherd is an Associate Professor in the Department of Computer Science at Louisiana State University.  He earned his Ph.D. and M.S. in Computer Science at the University of Delaware, and his B.S. in Computer Science at Virginia Commonwealth University.  David has since worked as a postdoctoral fellow in the Department of Computer Science at the University of British Columbia, built sweat equity as employee #9 at Tasktop Technologies, and risen to Senior Principal Scientist at ABB Corporate Research.  His research has produced tools that have been used by thousands, innovations that have been featured in the popular press, and practical ideas that have won business plan competitions.  Dr. Shepherd currently serves as the Co-Editor-in-Chief of the Journal of Systems & Software. His current work focuses on enabling end-user programming for industrial machines, increasing diversity in computer science, and using VR to treat ADHD.

05.10.2023 – The search for Knowledge—how technology changes everything you know (literally) and how you know everything

Speaker:  Dr. Daniel M. Russell

Host: Prof. Elaine Huang, Ph.D.

Abstract

In the internet age, how do people search for information?  Imagine your life without a search engine or without your phone-in-hand. Nearly instant access to the world’s knowledge has fundamentally changed the way we think about information and what knowledge really means.  I’ve studied how people search for information and knowledge online.  I know what they do with that information, and can tell you how to be better at doing this for you job or for your life.  At the same time, as AI and ML sweep across the news, what new realities can we expect from these technologies?  Dan Russell has been working at the intersection of AI and humans for 40 years and has seen AI tech come and go, and come back again. Surprisingly, AI is already deeply embedded in many of the products you use everyday.  Where will AI change products and our lives going forward?  It helps to see the future through the lens of the past—there’s hype, and then there’s reality. This talk will bring a uniquely deep perspective on the debates.

Bio

Dr. Daniel M. Russell has been working in Artificial Intelligence and Human-Computer Interaction for nearly 40 years. He has worked at several of the top technology invention companies in Silicon Valley (Google, Apple, Xerox, IBM) and has been at the forefront of many of their innovations.  He currently teaches in the Human-AI group at Stanford’s Computer Science department, and was in the core search engineering team at Google for over 17 years.  He has written over 200 technical articles for professional journals as well many articles for the popular press.  His most recent book, The Joy of Search: A Google Insider’s Guide to Going Beyond the Basics, is now out in paperback.  He has taught over 1000 classes in-person in venues ranging from 4th grade classes to professional classes for reference librarians at the Library of Congress.  His online classes have been watched by millions of students for an accumulated watch-time of over 400 years.

12.10.2023 – Current Trends in Human-AI Collaboration: Appropriate Reliance and Complementary Team Performance

Speaker:  Prof. Dr. Niklas Kühl

Host: Prof. Dr. Gerhard Schwabe

Abstract

As AI advice gains popularity in fields such as investment and medical treatment, effective human-AI collaboration becomes more important than ever. In this talk, I will elaborate on two key ingredients for such collaborations: appropriate reliance and complementary team performance. As AI is implemented in more and more high-stake decision making, decision-makers must discern how to "appropriately" rely on this often imperfect advice, accepting correct and rejecting incorrect suggestions. Despite its importance, current  research on this appropriate reliance lacks common definitions and operational measurements Thus, I will elaborate on the term and show first measurement results on how explanations for AI advice affect reliance behavior. Concurrently, I will elaborate on the ideal collaboration between humans and AI systems that should result in complementary team performance (CTP). This concept of CTP is underexplored, prompting a holistic theoretical foundation for understanding and developing human-AI complementarity, conceptualizing the potential and mechanisms of this collaboration. Both aspects focus on the nuances of human-AI collaboration, contribute to a broader understanding of how to effectively design AI advisors and leverage human-AI interactions to achieve levels of performance otherwise unattainable.

Bio

Niklas Kühl is Professor of Information Systems and Human-Centric Artificial Intelligence at the University of Bayreuth. He holds a senior management position at the Fraunhofer Institute for Applied Information Technology (FIT) and is Director at the FIM Research Institute for Information Management. Prior to his role in Bayreuth, he worked at IBM as a Managing Consultant in the area of Data Science and completed his PhD and habilitation at the Karlsruhe Institute of Technology. In his research, teaching and practical work, Niklas Kühl focuses on the interface of technical topics from artificial intelligence and machine learning with relevant problems from industry and society.

26.10.2023 – Kùzu Graph Database Management System

Speaker:  Prof. Semih Salihoğlu, Ph.D.

Host: Prof. Dr. Michael Böhlen

Abstract

In this talk, I will present the Kùzu graph database management system (GDBMS): an embeddable and feature-rich open-source DBMS that is optimized for ease of use, performance, and scalability. Datasets and workloads of popular applications that use GDBMSs require a set of storage and query processing features that relational DBMSs (RDBMSs) do not traditionally optimize for. These include optimizations for: (i) many-to-many (m-n) joins; (ii) cyclic joins; (iii) recursive joins; (iv) semi-structured data storage; and (v) support for universal resource identifiers. Kùzu aims to integrate state-of-art storage, indexing, and query processing techniques to highly optimize for this feature set. I will start by presenting the overall vision of Kùzu and what I think GDBMSs should offer beyond traditional relational DBMSs. Then, I will talk about several novel techniques we have integrated into Kùzu to be performant on m-n joins. Then I will given an example application  domain, graph machine learning libraries, that we are specifically integrating Kùzu with. Kùzu is actively being developed as an embeddable and fully functional open-source DBMS and is released under a permissible license.

Bio

Semih Salihoğlu is an Associate Professor and a David R. Cheriton Faculty Fellow at University of Waterloo. His research focuses on developing systems for managing, querying, or doing analytics on graph-structured data. His main on-going systems project is Kùzu, which is a new graph database management system that integrates novel storage, indexing and query processing techniques. He holds a PhD from Stanford University and is a recipient of the VLDB 2018 Best Paper, the VLDB 2022 Best Experiments and Analysis Paper award, and 2022 SIGMOD Research Highlights awards.

09.11.2023 – How and Why to Deal with Human Label Variation in NLP

Speaker:  Prof. Dr. Barbara Plank

Host: Prof. Dr. Rico Sennrich

Abstract

Human variation in labeling is typically considered noise. Annotation projects in computer vision and natural language processing typically aim at minimizing human label variation, to maximize data quality and in turn optimize and maximize machine learning metrics. However, variation in human labeling is ubiquitous, and the typical approach of minimizing human label variation by aggregation disregards human label variation. Recently, there is increasing evidence that human label variation is signal rather than noise [1,2] and critical for devising more human-facing, trustworthy language technology [3]. In this talk, I will first illustrate the problem and then discuss approaches to tackle this fundamental issue at the interplay of language resources, data quality, machine learning modeling and evaluation.

[1] Plank, B. (2022). The 'Problem' of Human Label Variation: On Ground Truth in Data, Modeling and Evaluation. In EMNLP. Association for Computational Linguistics.
[2] Basile, V. et al. (2021). We need to consider disagreement in evaluation. In Proceedings of the 1st Workshop on benchmarking: past, present and future (pp. 15-21). Association for Computational Linguistics.
[3] Baan, J., Aziz, W., Plank, B., & Fernandez, R. (2022). Stop measuring calibration when humans disagree. In EMNLP (2022). Association for Computational Linguistics.

Bio

Barbara Plank is Chair (Full Professor) for AI and Computational Linguistics, lead of the MaiNLP (Munich AI and NLP) research lab and Co-Director of the Center for Information and Language Processing (CIS) at LMU Munich. She is also professor at the Computer Science Department of the IT University of Copenhagen and ELLIS Scholar. She received her PhD in Computational Linguistics from the University of Groningen. Her research on tackling aspects of language variation within NLP, incl. data scarcity, multilinguality and human label variation is supported by the Munich Center for Machine Learning, a Danish Sapere Aude research leader grant from the Independent Research Fund Denmark and an ERC Consolidator grant.

23.11.2023 – From Automating Software Development to Empowering Software Developers

Speaker: Prof. Margaret-Anne Storey, Ph.D.

Host: Prof. Dr. Alberto Bacchelli

Abstract

Machines today can write software, compose music, create art, predict events, and listen and learn from humans. Notably, automation also plays an essential role in high performing software development teams by automating tasks and improving developer productivity.  But automation can’t (yet) replace human imagination and the intelligence that arises when multiple great minds work together to solve the complex problems that are inherent in software and systems design. In this talk, we will review how automation in modern software development has evolved and how it has disrupted how developers work. We will then explore how a deeper understanding of the developer experience points to untapped possibilities for innovating automation in software engineering, focusing on how how new innovations may:
1.    augment how developers may manage the cognitive complexity of today’s systems, 
2.    ease and enhance collaboration by speeding up feedback loops, and 
3.    help developers to get in and stay in a state of flow when developing.
We will conclude by discussing how we can measure the impact of new innovations on the developer experience, and how doing so will drive actionable change and empower developers to do their best work joyfully.

Bio

Dr. Margaret-Anne Storey is a Professor of Computer Science. She holds a Canada Research Chair in Human and Social Aspects of Software Engineering and is a member of the Royal Society of Canada’s College of New Scholars, Artists and Scientists. Together with her students and collaborators, she seeks to understand how software tools, communication media, data visualizations, and social theories can be leveraged to improve how software engineers and knowledge workers explore, understand, analyze and share complex information and knowledge. She has published widely on these topics and collaborates extensively with high-tech companies and non-profit organizations to ensure real-world applicability of her research contributions and tools. For the past several years, Dr. Storey consults with Microsoft and getDX to understand and improve developer productivity and developer experience. During the Fall of 2023, she holds the Inge Strauch Guest Professorship at the University of Zurich.

30.11.2023 – Transport Services: A new Standard API for the Internet

Speaker: Prof. Dr.-Ing. Michael Welzl

Host: Prof. Dr. Burkhard Stiller

Abstract

In recent years, the Internet's transport layer has undergone some significant changes. Notable examples of newly standardized Internet protocols are HTTP/3 (which uses QUIC instead of TCP) and Multi-Path TCP. Application developers should be able to benefit from the services that these new developments offer without having to rewrite their code. The Transport Services (TAPS) Working Group of the Internet Engineering Task Force (IETF) has just finished its work on a standard that makes this possible: TAPS replaces Berkeley sockets with a modern protocol-agnostic API, and it gives advice on how mechanisms underneath it can suitably engage and control network protocols based on application requirements.

One implementation of a TAPS system is particularly noteworthy: Apple's Network.framework. As one of the major contributors to the TAPS Working Group, Apple has been recommending this API to all their application developers and has been using it for its own applications since 2018 (beta for iOS 12 and macOS Mojave). Since 2022, an application using this API may benefit from QUIC without even noticing that it does not operate over TCP anymore.

This talk will give an overview of these new developments and discuss their implications. Besides providing more flexibility in the Operating System and the network, the richer service set of TAPS changes how we should think of networking, even when using higher-layer APIs (pub/sub middleware, REST, RPC, ..). This can create new opportunities for research on Distributed Systems and networked applications in general.

Bio

Prof. Dr. Michael Welzl is a full professor in the Department of Informatics of the University of Oslo since 2009. He received his PhD (with distinction) and his habilitation from the University of Darmstadt, Germany in 2002 and 2007, respectively, after receiving his MSc in Telekooperation at the Johannes Kepler University of Linz, Austria. Michael has been active in the IETF and IRTF for many years, e.g., by chairing the Internet Congestion Control Research Group (ICCRG) for 11 years, leading the effort to form the Transport Services (TAPS) Working Group, and authoring 13 RFCs. He has also participated in several European research projects in roles such as coordinator and technical manager. His research interests include transport protocols and network architectures, with a focus on the Internet and on performance as well as energy efficiency.

Click here for the archived IfI Colloquia pages.