My hope is that ML can be used to point out real world things you can't fake or work around, such as why an idea is considered novel or why the methodology isn't just gaming results or why the statistics was done wrong.