Stephen Downes

We often read that an AI, unlike a human, does not 'understand' what it is saying. But what is it even to 'understand'? There's no real consensus on this, and the best way to talk about it is to look at the varieties of understanding, for example as outlined in the book edited by the same author - here's the intro. In academia there's a sense that understanding amounts to being able to explain by means of a theory - this is known as 'objectual' understanding, but even here there are different ways of explaining - for example, do we limit our theory to causal explanation, or do we also take into account non-causal but environmental conditions for the possibility of the event (the kairetic criterion)? And then, there is what might be called first-order understanding, which does not involve theory or explanations, but is rather a more fundamental 'grasping' of the meaning and significance of something, as in 'I understand that murder is wrong'. Anyhow, this post was prompted by my reading this current paper on "non-factive objectual understanding' (NFOU)... the sort of understanding we have when we grasp the range of possible accounts of a phenomenon."