The next online seminar from the program Tübingen Science Bridge – Latin America (Humanities) will take place on July 19th at 04:00 pm MEZ and this year will focus on the Humanities and Social Sciences’ approach to Artificial Intelligence, a topic of crucial importance in the contemporary debate.

The lecture has the participation of Prof. Gabriela Arriagada Bruneau, Assistant Professor, Data Ethics and Artificial Intelligence Institute of Applied Ethics (IEA) / Institute for Mathematical and Computational Engineering (IMC) of the Pontificia Universidad Católica de Chile. She will put on the agenda the theme Title: “A bias network approach to articulate bias mitigation strategies in AI”.

The program, an initiative of the Baden-Württemberg Center for Brazil and Latin America at the Universität Tübingen, aims to contribute to the internationalization of science and research. Scientists from several partner institutions will present their latest research data, promoting an integrated and constructive environment for scientific interaction and contributing global knowledge. The lectures of the Tübingen Science Bridge are aimed at professors and scientific researchers, graduate students, as well as a more the general audience.

The online seminar will be held in English on the ZOOM platform in order to allow discussion and interaction.
Register link: https://tsb-july-23_2.eventbrite.de

More information about the Tübingen Science Bridge 2023 (agenda, lectures etc.): https://bit.ly/TSB-Agenda2023

Lecture Briefing:
In this presentation, I will talk about a work in progress proposal of a bias network approach to articulate bias mitigation strategies. Our proposal brings forward the idea of using a ‘bias network approach’ (BNA) to make decisions about bias mitigation strategies in algorithmic development, which is particularly useful for high social impact domains like healthcare. The BNA considers how biases in the AI pipeline can influence each other as well as their interaction with relevant societal and human structures (e.g., human decision-making processes, implicit biases, problematic artifacts, and material dependencies). Adopting a sociotechnical view of AI, we argue that using BNA as a reflective methodology can improve the quality of ethical assessments of algorithmic projects. In collaboration with a research team, we conducted a retrospective BNA analysis on their ML project that created a model used for healthcare waiting lists in Chile. We were able to identify intertwined ethical and technical challenges, biases, and decisions that were not evident at the time of the project’s execution. In addition to pointing out these weaknesses, this collaboration allowed the research team to envision new opportunities to improve their model for future upgrades by using the BNA.