Keynote - Ten Key Questions that a Company Should Ask to have Responsible AI

AI systems should be proportional to the problem they aim to solve - avoid using complex solutions when simpler ones would suffice

Data is only a proxy for reality and often fails to capture qualitative aspects - avoid over-relying on data-driven decisions without human judgment

Systems must allow for human contestability, auditability and the right to appeal automated decisions

Minimal data collection and storage time should be practiced following good data protection principles

AI systems should empower and augment human capabilities rather than fully replace humans

Bias cannot be fully removed but should be identified, measured and mitigated - transparency about limitations is crucial

Avoid creating fictitious categories or using arbitrary thresholds that oversimplify complex realities

Solutions should be technology-independent and regulate the use cases rather than specific technologies

Consider environmental impact - AI systems can have significant carbon footprints

Regular validation, testing and maintenance is required to ensure systems work as intended over time

Ethics and responsibility should be considered from the design phase, not as an afterthought

Digital divides and inequality can be amplified by AI - consider impacts on underserved populations

Allow for “I don’t know” responses rather than forcing predictions with low confidence

Explanations and transparency are crucial, especially for high-stakes decisions

Work with domain experts and ensure you have the right expertise for the problem you’re trying to solve