I don't think anybody is suggesting building an explicit "just say 'No' to extortion" heuristic into an AI. (I agree we do not have a good definition of "extortion" so when I use the word I use it in an intuitive sense.) We're trying to find a general decision theory that naturally ends up saying no to extortion (when it makes sense to).

Here's an argument that "saying 'no' to extortion doesn't win" can't be the full picture. Some people are more credibly resistant to extortion than others and as a result are less likely to be extorted. We want an AI that is credibly resistant to extortion, if such credibility is possible. Now if other players in the picture are intelligent enough, to the extent of being able to deduce our AI's decision algorithm, then isn't being "credibly resistant to extortion" the same as having a decision algorithm that actually says no to extortion?

ETA: Of course the concept of "credibility" breaks down a bit when all agents are reasoning this way. Which is why the problem is still unsolved!