I like the idea of energizing different letters proportionally to the model's probability & allowing users to land on different letters. That would lead to user

I like the idea of energizing different letters proportionally to the model's probability & allowing users to land on different letters. That would lead to users subconsciously affecting the outcomes in addition to the output from the model.