Learning in the Loop

Learning from user feedback well requires solving several problems simultaneously:\\ {a) Dealing with the scale of user feedback, which is often several orders of magnitude larger than editorial feedback.\\ {b) Dealing with the weak and sometimes adversarial feedback that users provide.\\ {c) Dealing with the shifting and online nature ofthese problems. I will discuss a combination of several tricks which address these problems effectively and some open questions that are bugging me.