Description
First we give a presentation about Ricks graduation thesis on the core principles of Responsible AI (i.e. fairness, explainability, etc), how these principles relate to each other and how one could apply these in code with the assistance of a process model Rick created to develop more responsibly. This process model helps one code AI more responsibly by posing questions related to these values. By answering these questions, one adheres to these principles and thus, in theory, would code AI more responsibly. This will be one of the topics of our presentation.One will find out in practice, however, that it's not always possible to take all principles into consideration evenly. In fact, more often than not you will find yourself in a situation where you have to deal with certain trade-offs between these principles in your project. Responsible AI is ethics, and in ethics one always ends up having to make a choice. In the context of Responsible AI for example, an AI-algorithm, more often than not, can't be completely transparent about its inner functioning and yet have the highest performance on solving a specific task in comparison with other models, especially large black-box models. A choice has to be made and this can be difficult.
Besides a small presentation about Ricks graduation project, we would like to talk about these trade-offs and how you could deal with them once they appear in your own projects. We will introduce four different trade-offs and would love to have an open discussion with your group on each of them with plenty of room for questions and interaction.
| Period | 23 Feb 2024 |
|---|---|
| Held at | Technische Universiteit Eindhoven , Netherlands |