I think that what you mean by "push" is somebody making a decision to call someone else's attention to an item.

The particular feature that you propose to test, however, a new "share" feature in Yahoo! Answers, is actually two features. First, it is a "raise your own topic and answer it" feature. Second, it is a "bring a (your) topic to the attention of someone else." But it seems to me that the pure test of your hypothesis would be to test a feature of bringing anything to the attention of someone else, without it having to be something that you yourself posted.

If there's only a reward for things that are recommended and liked, but no penalty for things that are not liked, you'll end up with spam. How about something where you get a penalty if the person doesn't like it?

After you show that there's some effect of cascading reward, perhaps you could try to assess how big the cascading reward needs to be to compensate for a reduction in the direct reward. For example, if the total payoff is the same, but it's distributed differently, what is the impact on motivation?

Any thoughts on how you prevent people from getting points by forwarding to their own shill accounts, or perhaps a cascade of them?

In general, I pretty much like this interesting incentive design and I bet it would be useful in some online situations, as it has been popular in some real markets:)
So I strongly encourage you to try it!

Here are a bit comments:
1, The introduction about "pushing" was a little bit confusing with the main design you are proposing. I don't think you need to focus on it too much. Just concentrate on your main ideas...since the cascading reward can be in some other situation than pushing.

2, Penalty would be useful to try for controlling spamming. However, as the major concern, this very explicit incentive and dis-incentive might make the contributors more of utilitarianism. And if penalty is possible, how would people like to try. as the common sense that when people recommending, people are doing good things and no reason to penalize

3, the new distribution of points might be a concern too. I think the dynamic will process in two directions depending on the initial conditions: either becomes communism or skewed pyramid; both can be dis encouraging.

4, as I mentioned in class, direct recommendation can be only useful within small scale: you only know what your friend wants and also it is ensured she will "like" what you recommend to her. So I think Yahoo answer might be not a good place for it.