Jazon Szabo of the 2019 cohort of the UKRI Centre for Doctoral Training in Safe and Trusted Artificial Intelligence contributed his paper, ‘Integrating Quantitative and Qualitative Reasoning for Value Alignment’ to the 19th European Conference on Multi-Agent Systems (EUMAS 2022).
Jazon and his co-authors’ paper focuses on the need to integrate qualitative and quantitative aspects of value-based reasoning. This is in response to the ‘value alignment problem’, where there is a need for machines to consider societal values alongside their own goals and actions.
As Jazon explains in his own words, “Values are central in teaching machines what we humans care about and so greatly influence how we would like them to behave. Values sometimes have more qualitative, human-like aspects: they can be vague, imprecise and approximate. On the other hand, values sometimes have more quantitative aspects: they may require complex calculations, which machines excel at. As a result, in this paper I’m looking at how to combine these two vastly different aspects in one mathematical formalism.”
The conference took place in Düsseldorf and was a great chance to build new connections with fellow academics. Jazon said, “I have found it great to meet other PhD students from around Europe. I have found it fun and useful.”
The paper is co-authored with Jose Such, Natalia Criado Pacheco and Sanjay Modgil