Normalizing the user preference vector is not necessary in this scheme. The magnitude of the preference vector is a constant scale on all of the comparisons to possible recommendable entities, so it drops out.
Overall, this is a reasonable starting point for recommending user actions.
When you have more data, you can try using collaborative filtering to estimate user preferences more robustly from sparse data. You could also try optimizing user and entity parameters by maximizing a likelihood function on followed recommendations composed of probabilities regressed from a function of the user and entity Mark vectors.
You're right about normalizing the user preference vector: not strictly necessary. Still if we're normalizing bounty vectors, it feels right to normalize everything :).
Another variation that I was considering is to generate 'user clusters'. In Mark Vector Space, divide user vectors into N groups such that the net variance across all clusters is minimized. Then when a user, for which there is sparse data, needs contextual information from other users, I could simply ask how correlated he is to the different clusters. If each cluster's 'center of mass' is a vector, the dot product between a new user and the different cluster vectors could be informative in reconstructing suggestions: the idea being to infer from similar users what a particular user might want.
I was also wondering whether adding a stochastic component to each user-vector would be interesting.
Overall, this is a reasonable starting point for recommending user actions.
When you have more data, you can try using collaborative filtering to estimate user preferences more robustly from sparse data. You could also try optimizing user and entity parameters by maximizing a likelihood function on followed recommendations composed of probabilities regressed from a function of the user and entity Mark vectors.