- Cet évènement est passé
AI4EU Web Café : Christèle Tarnec (Orange Labs) with her talk on: « Ethics by design in AI: Guidelines and toolkits to develop fair & explainable models » (15h00)
15 octobre 2020 de 15 h 00 min
On October 15th, 2020, at 3:00 pm with the inspiring Speaker :
Christèle Tarnec (Orange Labs)
with her talk on : « Ethics by design in AI: Guidelines and toolkits to develop fair & explainable models »
What is her talk about:
I will share with you an analysis of existing tools and toolkits to help data scientists design fair and explainable AI models. A particular focus will be done on IBM toolkit AIF360 (70 metrics and 11 bias mitigation algorithms) and AIX360 (9 algorithm to explain AI models). This work has been done with Hugo Miralles (a student from UPC School Barcelona/Ecole Centrale de Nantes-France).
I have been working in Orange Labs for 17 years. With a business school background, I have been involved in many projects (IoT, smart cities, Data management platform, Ethics by design in AI). I always have an Innovation Management approach that considers technologies will enhance services for users. With fellow engineers, we have been studying for 18 months the ethical issues that any data scientist can face during the construction of AI services and how to start using ethics by design methodology to support them. In spring 2020 with two colleagues among whom a student from UPC School Barcelona/Ecole Centrale de Nantes-France, we studied the different open source tools available to develop fair and explainable models. IBM toolkit offers the most functionalities and we studied it in detail. This web café will be an opportunity to share with you our analysis, exchange on what remains to be done and share your practices.