How to add (soft?) constraints to the model?

Questions
3

The function is sigmoid. I wonder how to incorporate this additional information I have to help with the learning… I’m not sure if this could be integrated into the prior either