Keynote Speakers

Daniel Gatica-Perez

Daniel Gatica-Perez directs the Social Computing Group at Idiap and is a professor at EPFL in Switzerland. His research integrates theories and methods from ubiquitous computing, social media, and social sciences to understand human behavior for social good applications. His research has been supported by the Swiss National Science Foundation, the European Commission, and several industry partners. He has also worked with cities and local organizations in social innovation projects. As a member of the ICMI community, he served as Program Co-Chair in 2009 and 2011, and as chair of the Steering Board between 2013 and 2015.

Title: Towards Increasing Diversity in Datasets

Abstract: Diversity is fundamental for the design of technology that serves people and is relevant beyond academic pursuit. In the talk, I will share experiences from our work related to diversity in data, and highlight the importance of surfacing the values that guide our research: people at the center, community building, and multidisciplinarity.

Laura Cabrera-Quirós

Laura Cabrera-Quirós is an assistant professor at the Instituto Tecnológico de Costa Rica (TEC), Costa Rica. She received her PhD from the Delft University of Technology (TU Delft) in the Netherlands, and has previous experience as an electrical engineer working with embedded systems. Before joining TEC, she worked as a postdoctoral researcher at TU Delft and at Eindhoven University of Technology (TU Eindhoven) in the Netherlands, in projects collaborating with the Maxima Medical Center and Philips research. Her main research interest is to leverage the fusion of machine learning with the use of embedded systems (e.g. wearables and camera technologies) for the acquisition of data and analysis of human signals in a non-invasive manner, in order to understand human behavior, monitor health, and in general improve people’s life quality.

Title: The MatchnMingle dataset: some untold stories and challenges behind collecting and annotating data from real social situations in the wild

Abstract: Every paper introducing a novel dataset or annotation tool has very valuable information about the resource itself and how to use it for further work. But only a few of these give a glimpse into all the efforts and sometimes inherent nuisances and challenges that are encountered by the research teams while collecting and annotating data. Some of these challenges increase when dealing with real social data. As greater need to create new datasets to answer specific research questions arise, it is important to share these generally untold experiences amongst the community. Doing so could reduce the potential time spend by teams working on challenges that other teams have previously faced and tackled, and could also benefit reproducibility. This keynote intends to use the process of collecting and annotating the MatchMingle dataset as a case study, to open the discussion about some relevant but generally untold stories.

Tobias Baur

Tobias Baur is a post-doctoral researcher at the Human-Centered Artifical Intelligence Laboratory, Augsburg University, Germany. His main research focuses on human-centered tools that help (non-)experts to create and understand Machine Learning models. His research topics include Artificial Emotional Intelligence; Social Signal Processing; Machine Learning; Explainable AI; Affective Computing and Human Computer Interaction.

Title: eXplainable Cooperative Machine Learning – An interactive Human/Machine Annotation Process

Abstract: In this keynote we introduce a novel annotation workflow, which we subsume under the term “eXplainable Cooperative Machine Learning” and show its practical application in a data annotation and model training tool called NOVA. The main idea of this approach is to support human annotators already during the annotation process with AI techniques. This allows annotators to incorporate semi-supervised active learning techniques in their workflow by providing the possibility to pre-label data in a semi-automated way, resulting in a drastic acceleration of the annotation process. A highlight of this approach is that it gives the annotator full control over the training process and insights into their AI models decisions. To achieve this, we combine this workflow with recent eXplainable AI techniques.