Objective:
Predict a(single) real valued score for userA --- userB given a set of parameters
Queries:
Is it commutative? score(userA, userB) = score(userB, userA)
The real concern in such systems is the way to figure out the best ranking either subjectively or as you said mathematically, which can be manipulated with strong prior probabilities. For example, in case where strongly agree, agree etc 3 of 6 are chosen, there's a joint probability distribution of 3 of the 6 options to be true.
Priority:
The concern regarding this can be easily sorted out with weighted ranking from high to low as we go down the priority scale.
Options:
A statistical measure called entropy and correlation can be computed easily between any 2 users, given a set of attributes ranging from selection on Likert scale, priority order, and others.
OR
Conversely a scoring system can be learnt from data at hand that is subjectively graded apriori.
About me:
Being a research student in machine learning and statistics, I find this very interesting. The funds accumulated from taking up such projects go into conference and publications.
Kindly message me whenever you're ready. If not online, leave your queries, and I shall reply ASAP.
Thank you! Have a great weekend!