Maybe. But there is a better way to evaluate it. Let's say that adding recognition of human player's wrong answers would take an extra 5% of development effort and an extra 5% hardware and software cost. I think both would be a lot more but for the sake of argument lets use those figures.
The question then becomes is it worth spending another 5% development cost and 5% HW/SW cost for this capability? What does it really give you? The ability to rule out one out of several possible answers from the set of answers Watson generates. We can actually empirically evaluate how useful that would be. The only time it would be useful would be in use cases where Watson gave the correct answer that a human player already gave. As I understand it that's happened once in several games. Its not worth the cost even if we accept that its relatively trivial to implement. About Smarter Ideas Read the Article at HuffingtonPost