25.11.2021 | Inside Data Science
Hedrik Heuer über Machine Learning und Mensch-Computer-Interaktion
Hendrik Heuer, Träger des Weizenbaum-Studienpreises 2021, spricht über die Bedeutung von Daten in seiner Forschung und gibt Ausblick auf eine neue Definition für Machine Learning.
What topics are you currently working on in your research?
I’m a computer scientist working at the intersection of machine learning and human-computer interaction. In my dissertation, I provide a socio-technical perspective on ML-based curation systems like YouTube’s recommendation system. My thesis examines how users and practitioners make sense of such ML systems. In the thesis, I provide actionable insights on how ML-based curation systems can and should be explained and audited. As a Postdoc, I now examine how algorithmic regimes shape science. I also develop tools against disinformation and leverage ML techniques to help people with disabilities.
How important is data to your research?
For me, machine learning is a novel programming paradigm in which decisions are inferred from data. In almost all of my work, I either study or apply such machine learning. Therefore, data is central to my research. In an upcoming paper, we will even argue why data-based automation is a much more precise term than machine learning because neither the term machine nor the term learning really captures how ML algorithms leverage data to infer decision rules.
What role does data science play in your research? Do you see yourself more as a user, a method developer, a basic researcher, or perhaps something completely different?
I see myself primarily as an HCI researcher. As the title of my dissertation “Users & Machine Learning-based Curation Systems” illustrates, I investigate the whole socio-technical system. On the technical side, this includes input data, ML algorithms, inferred models, and output data. On the social side, I study providers of data, ML practitioners, auditors, the organizations that operate a system as well as the primary, secondary, and tertiary users of a system. Im inspired by Adrian Mackenzies wonderful book “Machine Learners – Archeology of a Data Practice”.
Which data science methods and technologies are in the focus of your research or could also become interesting in the future?
In the past, I have been primarily interested in supervised machine learning. Before I started my doctoral studies in Bremen, I worked as a deep learning researcher in Amsterdam. Thanks to a grant awarded by the Data Science Center, I will be able to apply data science techniques to automatically recognize text in images and to summarize this text for people with disabilities. I am also interested in using these techniques to help people recognize misinformation and fake news. I do not necessarily think that such tasks can be fully automated, but I think better tools to search and analyze text are needed.
What are your main challenges in dealing with data?
The biggest problem is finding enough time for all my ideas. Apart from that, obtaining high-quality labeled data is a big challenge.
And finally, what is your personal motivation for joining the Data Science Center?
I had the pleasure of working closely with my colleagues in the CRC 1342 on WeSIS, the Welfare State Information System. Our goal is to empower social scientists to use machine learning and other data science techniques to their advantage. During that time, I saw how much exciting work in this space is already happening in Bremen. The Data Science Center is the place where all these researchers come together to discuss data science and machine learning.
Mehr zu Hendrik Heuers Tätigkeiten erfahren Sie am 02.12.2021 in seinem Beitrag „A Socio-Technical Perspective on the Potentials and Perils of Machine Learning“ im Data Science Forum.
Dr. Hendrik Heuer
Postdoc @ Information Management AG & ifib
FB 03 – Mathematik und Informatik