How do you handle a situation where the data contradicts your initial assumptions and the domain experts disagree with your model?
