Inspired by a design presented at VoiceLunch US/Canada, Mark and Allen discuss how we would implement the design - consisting of keeping a "score" about user's to our skill or action, adjusting the score sometimes, and presenting different prompts based on this score. While the design seems straightforward, there are a number of interesting approaches  to the implementation.  

How would you implement this design?  

What libraries do you want to see to help make your implementations easier?