Extrapolated CEV would be working from observable evidence + a good prior. Whereas lots of people insist it's very important to them to believe in a deity through faith, despite any contrary evidence (let alone lack of evidence). How are you going to tell the CEV to ignore such values?