Rank Aggregation in Social Choice and Machine Learning

Humans are notoriously bad at rating things "on a scale from 1 to 10" , and yet we often ask them to do so when getting information about their preferences, from Amazon to Netflix to rate my professor.com. We can do a lot of interesting theory on discretized ratings by working on the real line and then quantizing, but if the input from a human isn't accurate in the first place, that doesn't do a lot of good. I'll tell you about "ranking" or "pairwise comparison" data where we ask people to compare items, which is often more natural than rating them. I'll discuss some interesting old work in social choice theory, about voting in elections, and then tell you about the work done on this area in machine learning.