A very useful distinction between evaluating whether an AI produces an acceptable answer and evaluating the competence behind that behaviour.
There may be a further governance layer worth considering. Even if moral competence can be evaluated reliably, organisations still need to determine which evidence was considered, which contextual boundaries applied, who had authority to accept or override the system’s recommendation, and whether that decision can later be reconstructed.
In other words, moral competence may help us evaluate the model; traceable human authority may help us govern its use.
The combination seems especially important as AI moves from answering questions to participating in consequential decisions.
Great Stuff, very thorough!
Thank you, Leonardo!
A very useful distinction between evaluating whether an AI produces an acceptable answer and evaluating the competence behind that behaviour.
There may be a further governance layer worth considering. Even if moral competence can be evaluated reliably, organisations still need to determine which evidence was considered, which contextual boundaries applied, who had authority to accept or override the system’s recommendation, and whether that decision can later be reconstructed.
In other words, moral competence may help us evaluate the model; traceable human authority may help us govern its use.
The combination seems especially important as AI moves from answering questions to participating in consequential decisions.