If our aim is to build morality into an artificial agent, how might we begin to go about doing so?

Moral disagreement is fundamental: The paper identifies that "what it means to be moral is still debated," highlighting that there is no universal consensus on morality itself—a critical challenge when attempting to encode it into AI systems for high-stakes decisions in healthcare and autonomous driving.

Both design approaches have limitations: The paper discusses "top-down and bottom-up approaches to design," suggesting that neither method alone is sufficient; a hybrid approach combining both is proposed as more promising for building moral agents.

A single moral framework is insufficient: The paper proposes "a hierarchical approach to combining moral paradigms" rather than relying on one ethical theory, recognizing that different moral frameworks (like utilitarianism or deontology) may be needed for different contexts.

Governance and policy are essential safeguards: The paper emphasizes that "governance and policy are becoming ever more critical in AI Ethics," indicating that technical solutions alone cannot ensure moral AI—external oversight and regulation are necessary to verify that systems actually behave ethically.

We must ensure tasks are actually achievable: The paper stresses the importance of "ensuring that the tasks we set for moral agents are attainable," warning against programming AI systems with impossible or contradictory moral objectives that cannot realistically be fulfilled.