Implementing Ethical AI. An Ethics First Approach to the Development of AI Systems

Machine Learning & Künstliche Intelligenz | Vortrag | 60 min | Englisch

MO | 12:00 | BAHNHOF ST. PAULI

Nothing is perfect. Even algorithms can make mistakes. They can be biased. We shall present a procedural framework for AI systems’ development that acknowledges these risks. Our framework highlights the ethical implication and consequences throughout the development cycle and provides a set of mitigation strategies. Building on the Disciplined Agile Delivery framework, our approach includes two additional Ongoing Process Goals when developing AI systems: ethical threat modelling and social impact assessment. The former emphasises questions related to transparency, effectiveness and the systemic scope. The latter provides a set of tools to empirically assess whether groups are treated differently by the AI system. By systematically prioritising ethical matters together with security, technical or business concerns, we create awareness about the social impact of the systems that we build. Only then can we create human-centered AI systems.

Matthias Orlowski

areto Consulting GmbH

Matthias Orlowski works as a data scientist at areto consulting gmbh. Since January 2017 he has been responsible for the data science competence center there. After completing his PhD in 2015, Matthias has been working on various projects in the fields of consumer targeting, campaigning and people analytics.

 

Anna Lundemo

areto Consulting GmbH

Anna Lundemo joined areto consulting gmbh as a data scientist in 2018. She focuses on the development of Natural Language Processing solutions for people applications.