Change the scoring model to enable "hand raising"

[morsh]

This will enable encapsulating each model in its own logic where each model flags it's ability to receive an utterance.