Program

• After you register for EACL 2021, see the workshop page within virtual2021.eacl.org for the Zoom and GatherTown links.

• The poster and discussion sections will be held in poster room 2.

• We will also post announcements on the RocketChat channel there. (Direct link)

• For posting or following on Twitter, please use the hashtag #hcinlp.

08:00am - 08:15am ET / 14:00 - 14:15 CEST Welcome (Zoom)

8:15 -9:00 ET / 14:15 - 15:00 CEST Invited Talk: Burr Settles (Zoom)

Making Natural Language Processing More... "Natural"

Abstract: Language is a fundamental part of human communication. It's curious, then, that so little research in machine learning and NLP involves conducting human user studies, and in fact many assumptions we make in benchmark or simulation studies simply don't hold with humans are in the loop. In this talk, I'll draw on a range of past projects — from smart annotation interfaces for active learning in NLP to AI-driven user experiences in Duolingo, the world's largest language-learning platform — to illustrate not only that user studies are often key to driving real improvements, but considering the end user experience carefully can drive innovative new NLP applications as well.

Bio: Burr Settles leads the research group at Duolingo, an award-winning website and mobile app offering free language education for the world. He also runs FAWM.ORG, a global annual songwriting experiment. He is the author of Active Learning — a text on machine learning algorithms that are adaptive, curious, and exploratory (if you will). His research has been published in CogSci, NeurIPS, ICML, AAAI, KDD, ACL, EMNLP, NAACL-HLT, and CHI, and has been covered by The New York Times, Slate, Forbes, WIRED, and the BBC among others. In past lives, he was a postdoc at Carnegie Mellon and earned a PhD from UW-Madison.

9:00 - 9:45 ET / 15:00 - 15:45 CEST Lightning Talks (Zoom)

9:45 - 10:00 ET / 15:45 - 16:00 CEST Break

10:00 - 11:00 ET / 16:00 - 17:00 CEST Poster Session #1 (Gather, Poster Room 2)

11:00 - 12:00 ET / 17:00 - 18:00 CEST Round Table Discussion #1 (Gather, Poster Room 2)

12:00 - 12:30 ET / 18:00 - 18:30 CEST Break

12:30 - 13:15 ET / 18:30 - 19:15 CEST Invited Talk: Marti Hearst (Zoom)

Putting the H into NLP

Abstract: Much of my research over the last 30 years has been at the intersection of HCI, NLP, and search. In this talk, I will share tips for designing user interfaces that make use of NLP, and tips for evaluating NLP algorithms that make use of HCI methods. I will also talk about some recent work that combines these two areas, including the ScholarPhi project in which we turn explainable AI on its head and use NLP to better explain AI papers, a new text simplification algorithm that uses ideas from HCI to assess its behavior in a novel way, and an NLP-based semantic improvement to the pernicious word cloud visualization.

Bio: Marti Hearst is a Professor at UC Berkeley in the School of Information and the Computer Science Division. Her research encompasses user interfaces with a focus on search, information visualization with a focus on text, computational linguistics, and education at scale. She is the author of Search User Interfaces, the first academic book on that topic. She co-founded the ACM Learning@Scale conference. She is a former President of the Association for Computational Linguistics, a member of the CHI Academy and the SIGIR Academy, an ACM Fellow, and has received four Excellence in Teaching Awards from the students of UC Berkeley.

13:15 - 14:00 ET / 19:15 - 20:00 CEST Poster Session #2 (Gather, Poster Room 2)

14:00 - 14:15 ET / 20:00 - 20:15 CEST Break

14:15 - 15:15 ET / 20:15 - 21:15 CEST Round Table Discussion #2 (Gather, Poster Room 2)

15:15 - 16:00 ET / 21:15 - 22:00 CEST Invited Talk: Niloufar Salehi (Zoom)

How algorithmic personas can help people shape algorithmic systems

Abstract: In this talk I will discuss how we might engage people in shaping algorithmic systems that are often complex and difficult to discuss. I argue that rather than engaging potential users and community members on how the algorithm works, we will be better off developing methodologies to engage people on what the algorithm does. In other words, how might we design algorithmic systems by centering what end-users perceive as the social role that the algorithm will play. Machine learning systems are usually developed to predict the most likely outcome based on training data, effectively mimicking the types of decisions that people might make. For instance, a translator takes text in the source language and provides the closest text in the target language, and a doctor determines whether a digital pathology image is cancerous or not. But users have difficulty adopting these tools if they do not understand its capabilities, its intended use, or what it might be able to do differently. I argue that if we start from the point of understanding the potential social role that the algorithmic system will play, we can develop alternative framings for what users expect the system to do, what information and interactions they need to use it effectively and reliably, and how they might be able to help shape the system. I will discuss methodological advances in studying and prototyping algorithms using algorithmic personas and novel algorithmic prototypes.

Bio: Niloufar Salehi is an Assistant Professor at the School of Information at UC, Berkeley, with an affiliated appointment in EECS. Her research interests are in social computing, participatory and critical design, human-centered AI, and more broadly, human-computer-interaction (HCI). Her work has been published and received awards in premier venues in HCI including ACM CHI and CSCW. Through building computational social systems in collaboration with existing communities, controlled experiments, and ethnographic fieldwork, her research contributes to the design of alternative social configurations online.

16:00 - 16:10 ET / 22:00 - 22:10 CEST Closing (Zoom)