ETH continuing education course addresses ethics in AI

What is philosophy? What is ethics in the era of artificial intelligence (AI)? These questions are discussed in small groups by 20 students taking the CAS ETH Machine Learning in Finance and Insurance, a continuing education and training programme that is being held for the first time this year. The course is based at ETH Zurich’s FinsureTech Hub, as part of the professorships of Josef Teichmann and Patrick Cheridito at the Department of Mathematics.

From the outset, it has been clear to Programme Manager Bastian Bergmann that particular importance must be placed on the responsible use of AI technology when it comes to machine learning in a continuing education programme. “For me, ethics is just as important as the technology itself. How is the technology implemented in the context, and what challenges does it present? These are important questions that we deal with in the course,” says Bergmann, who is a financial mathematician and philosopher himself.

Andrea Ferrario, another lecturer on the course, shares this view: “Ethics doesn’t just become relevant when an artificial intelligence-based application arrives in society. Rather, the ethical dimension must be taken into account throughout design – from the outset and in every single step.” The physicist obtained a doctorate in mathematics at ETH and now researches biomedical ethics and the philosophy of artificial intelligence at the University of Zurich.

Developing your own perspective for the first time

For the two lecturers, the priority is that the CAS students, who all work in the Swiss finance or insurance sector, can go back to their company and build bridges between data scientists, software engineers, business experts, and colleagues from the compliance department who deal with legal requirements of artificial intelligence-based services and products. Accordingly, interdisciplinary discourse is particularly important to both lecturers. “Collaboration between the various professional groups, with their different approaches, opinions and objectives, isn’t always easy in practice,” says Ferrario. For him, the course is a valuable test bed for discussing ethical questions in the financial and insurance sector.

During the programme, Bergmann has repeatedly found that the subject of ethics is uncharted territory for most participants: “They're more used to monitoring systems that are based on a predefined set of rules. In the ethics part of the CAS, they’ve had to develop their own point of view. What moral implications does this have for me and my company?”

AI hallucinations and reputational risks

The two external lecturers Christiane Hoppe-Oehl, Head of the Artificial Intelligence Unit at FINMA, and Luca Baldassarre, AI Governance Director at Swiss Re, also noticed that discussing ethical topics was largely unfamiliar territory for the students. Both lecturers welcome the fact that ETH places so much importance on the ethical dimension.

“My key message to the students was that companies without a responsible approach to planning, developing and monitoring AI systems are exposed not only to business risks but also to risks in terms of compliance and reputation,” says Baldassarre. Hoppe-Oehl adds: “AI has the potential to drive advances in society, but if the risks are not recognised and dealt with, we might lose control.” She mentions the example of AI hallucinations, which can hamper informed decision-making because it’s more difficult to distinguish facts from untruths.

For Xenia Tao, a student on the programme, the ethical dimension of the CAS was one reason why she chose to deepen her knowledge of AI at ETH. “I find the integration of ethical considerations with real-world applications particularly valuable, especially as we navigate the evolving role of technology in finance,” says a former Managing Director at a Swiss bank, now focused on advancing ethical AI governance in the financial sector. Her fellow student Arantzazu Garcia also found the mixture of technical principles, ethics and practical orientation to be a particularly appealing feature of the CAS. “By working on a project of my own, I’ve been able to apply the concepts we’ve learned directly,” says the Global Operational Excellence Lead at Hitachi Energy.

Concluding with a valuable practical component

The practical component of the programme centred on projects and workshops with industry partners. Using specific examples, the students grappled with the implementation of AI applications in finance and insurance. Now, the students are working on their own projects in the final stage of the CAS. So far, Bergmann is pleased with how the first course has gone. It has proven worthwhile to start with the technical principles of machine learning and to move on via ethics to reach practical applications. So, will everything be the same next year? “Almost,” says the Programme Manager. “We still want to expand the ethics module a bit.”