Doctoral and Postdoctoral Researcher: Design of Trustworthy Artificial Intelligence

If you are interested in one of the topics, please apply through the application portal.

The professorship aims to develop principles of trustworthy artificial intelligence, implement them technically, and test them together with users in practical applications. 

RS1 – Trust in AI Through Understanding: 

This research specialization aims to measure and analyze the users’ understanding of AI systems (User Beliefs, Folk Theories). This specialization is motivated by our prior work on user beliefs in the context of machine learning-based video recommendation systems. This prior work will be extended and intensified (Alvarado et al., 2020), especially for the transparency and explainability of AI systems (Heuer, 2021). So far, we focused on YouTube. In the future, we want to study other platforms like TikTok and ChatGPT and novel social media platforms.

RS2 – Trust in AI Through Control:

This research specialization will develop methods for controlling the output of ML-based systems through systematic audits. An important motivation of this research specialization is to ensure non-discrimination. This specialization is motivated by prior work on online radicalization on YouTube (Heuer et al., 2021a). Important considerations include how such audits could enable a kind of TÜV or Stiftung Warentest for machine learning systems (Heuer, 2020).

RS3 – Trust in AI Through Co-Creation:

The third research specialization aims to develop methods for users’ deep involvement in designing AI systems. Our focus on such deep user involvement is informed by our experience developing the WeSIS information system (Molina León et al., 2022), where we deeply involved social scientists in interdisciplinary collaborations. We have also already completed an extensive literature review on the different approaches to user engagement in general (Jarke et al., 2021) and in the context of natural language processing (Heuer and Buschek, 2021). Our primary motivation is democratizing machine learning techniques (Heuer et al., 2021b).

The professorship follows the tradition of participatory and user-centered software development (MacKenzie, 2013) and extends this tradition with a focus on machine learning (Buschek et al., 2021). Participatory software development has a long tradition (Floyd et al., 1989) and offers numerous methods of actively involving users in software development (Bødker et al., 2022). Our goal is to develop these methods further with regard to artificial intelligence and machine learning as a new kind of co-creation (Bovaird, 2007; Clement et al., 2012; Gomillion, 2013; Jarke, 2021). 

One application context in which the research program will explore human-AI interaction methods is social media. This context is particularly well suited because AI systems are central to platforms like Facebook, YouTube, TikTok, Instagram, and Twitter. On YouTube, 500 hours of content are uploaded every minute (Hale, 2019). Artificial intelligence is needed to curate this large number of content and select the right videos for individual users. Up to 70% of videos viewed on YouTube were chosen by an AI system (Solsman, 2018). Research on human-AI interaction in this application field is socially relevant because users generally have no control over AI systems (Heuer, 2020) and often are unaware of the systems’ existence (DeVito, 2017; Eslami et al., 2015).

Social media is socially relevant because many people use it regularly. According to the Reuters Institute Digital News Report, WhatsApp is Germany’s most popular social media platform. In second place is YouTube (52%), followed by Facebook (41%) and Instagram (28%) (Newman et al., 2022). Companies from the USA or China operate most of the social media platforms. The platforms are funded through advertising revenue and surveillance (Zuboff, 2019). Right now, opportunities for co-creation are limited.

If you want to join us in working on these crucial issues, please apply through the application portal.

References

• Alvarado, O., Heuer, H., Vanden Abeele, V., Breiter, A., Verbert, K., 2020. Middle-Aged Video Consumers’ Beliefs About Algorithmic Recommendations on YouTube. Proc. ACM Hum.-Comput. Interact. 4. https://doi.org/10.1145/3415192
• Bødker, S., Dindler, C., Iversen, O.S., Smith, R.C., 2022. Participatory Design, Synthesis Lectures on Human-Centered Informatics (SLHCI). Springer.
• Bovaird, T., 2007. Beyond engagement and participation: User and community coproduction of public services. Public administration review 67, 846–860.
• Buschek, D., Loepp, B., Hauptmann, H., Wörndl, W., 2021. Ucai 2021: Workshop on user-centered artificial intelligence. Mensch und Computer 2021-Workshopband.
• Clement, A., McPhail, B., Smith, K.L., Ferenbok, J., 2012. Probing, mocking and prototyping: participatory approaches to identity infrastructuring, in: Proceedings of the 12th Participatory Design Conference: Research Papers-Volume 1. ACM, pp. 21–30.
• DeVito, M.A., 2017. From editors to algorithms: A values-based approach to understanding story selection in the Facebook news feed. Digital Journalism 5, 753–773.
• Eslami, M., Rickman, A., Vaccaro, K., Aleyasen, A., Vuong, A., Karahalios, K., Hamilton, K., Sandvig, C., 2015. “I Always Assumed That I Wasn’T Really That Close to [Her]”: Reasoning About Invisible Algorithms in News Feeds, in: Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems, CHI ’15. ACM, New York, NY, USA, pp. 153–162. https://doi.org/10.1145/2702123.2702556
• Floyd, C., Reisin, F.-M., Schmidt, G., 1989. STEPS to software development with users, in: European Software Engineering Conference. Springer, pp. 48–64.
• Gomillion, D., 2013. The Co-Creation of Information Systems. ProQuest LLC.
• Heuer, H., 2021. The Explanatory Gap in Algorithmic News Curation, in: Multidisciplinary International Symposium on Disinformation in Open Online Media. Springer, pp. 1–15.
• Heuer, H., 2020. Users & Machine Learning-Based Curation Systems (PhD Thesis). University of Bremen, Germany.
• Heuer, H., Buschek, D., 2021. Methods for the Design and Evaluation of HCI+ NLP Systems, in: EACL 2021 Workshop on Bridging Human-Computer Interaction and Natural Language Processing (HCI+NLP Workshop).
• Heuer, H., Hoch, H., Breiter, A., Theocharis, Y., 2021a. Auditing the Biases Enacted by YouTube for Political Topics in Germany, in: Mensch Und Computer ’21.
• Heuer, H., Jarke, J., Breiter, A., 2021b. Machine learning in tutorials – Universal applicability, underinformed application, and other misconceptions. Big Data & Society 8, 20539517211017590. https://doi.org/10.1177/20539517211017593
• James Hale, 2019. More Than 500 Hours Of Content Are Now Being Uploaded To YouTube Every Minute. Tubefilter. URL https://www.tubefilter.com/2019/05/07/number-hours-video-uploaded-to-youtube-per-minute/ (accessed 1.30.20).
• Jarke, J., 2021. Co-creating Digital Public Services for an Ageing Society : Evidence for User-centric Design. Springer Nature. https://doi.org/10.1007/978-3-030-52873-7
• Jarke, J., León, G.M., Zakharova, I., Heuer, H., Gerhard, U., 2021. Beyond Participation: A Review of Co-Creation in Computing. arXiv preprint arXiv:2111.04524.
• MacKenzie, I.S., 2013. Human-Computer Interaction: An Empirical Research Perspective, 1st ed. Morgan Kaufmann Publishers Inc., San Francisco, CA, USA.
• Molina León, G., Skitalinskaya, G., Düpont, N., Klaff, J., Schlegel, A., Heuer, H., Breiter, A., 2022. Co-Creating a Research Data Infrastructure with Social Policy Researchers, in: Proceedings of 20th European Conference on Computer-Supported Cooperative Work. European Society for Socially Embedded Technologies (EUSSET).
• Newman, N., Fletcher, R., Robertson, C.T., Eddy, K., Nielsen, R.K., 2022. Reuters Institute Digital News Report 2022.
• Solsman, J.E., 2018. YouTube’s AI is the puppet master over most of what you watch.
• Zuboff, S., 2019. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power: Barack Obama’s Books of 2019. Profile.