On Thursday, April 25, MDSI in collaboration with TUM Think Tank offers a special PhD training on "A Human Rights-based Approach to AI Development". Our coaches will be joining us from the < A+ > Alliance for Inclusive Algorithms and the womanatthetable initiative.

About the workshop

It is a common belief in the engineering and data science community that data presents neutrality and truth. In this course, we unpack this myth and demonstrate that data and AI systems are both relative and contextual. It dives into the mechanisms and culture around AI development that can carry and transfer biases and inequalities into AI systems.Through a human rights-based approach, this course equips its participants with the tools and common vocabulary to create technology for the dignity, equality and worth of humans. Focusing on the impact of the technology on human beings and their fundamental, inalienable rights, we relate human rights aspects to the entire AI life cycle.

The course aims at increasing the awareness of the interconnectedness of human rights principles and computing technology so that technology creators and deployers can actively follow human rights values rather than unwittingly harm them.

Structure

This one-day in person training is divided into the following four content modules:

1. Human Rights & AI Systems
   1. What are Human Rights? Core principles & legal frameworks
   2. How AI systems can contradict the core values of Human Rights
2. How Harms to Human Rights Enter the AI Lifecycle
   1. Introducing the six stages of the AI lifecycle
   2. For  each stage:
      1. Outlining the purpose of the stage
      2. Demonstrating different entry points of bias
3. Fairness Metrics: Technical Measures are Not Neutral
   1. Using the example of fairness metrics to illustrate that different metrics lead to different, i.e. contradictory real-world outcomes. Which metric is the most suitable for a system has to be a conscious choice, informed by the system’s socio-technical context.
4. A Human Rights-Based Approach to AI Development
   1. For each stage of the life cycle:
      1. Introducing considerations and reflection points that are required to create Human Rights-respecting AI systems.
      2. Best practice examples or tools that support the practical implementation
   2. Discussing two case studies to demonstrate that a Human Rights-based approach can manifest in very different measures depending on a system’s objective - even for the same dataset and functionality.

The content of each module is illustrated with real-life examples. Frequent slots for discussions and interactive exercises prompt the students to critically reflect and transfer the learned content.

Time & Place

The workshop takes place at MDSI (Blue Room, 5th Floor, Walther-von-Dyck-Str. 10, 85748 Garching)

Time is 9am to 5pm.

Sign Up

If you are interested in joining the workshop, please write a short email to info@mdsi.tum.de and state your interest. We need your name and affiliation, so if this is not visible from your email address, please share that information.