Netpref Presents

The Netflix Recommendation System

How does it work?

Lisa Singh: There are a number of different technologies that are used in recommendation systems. Some of them focus on understanding user preferences by looking at other users who have similar types of backgrounds or behaviors online. Netflix uses a number of different types of recommendation systems over the years. I’m not sure exactly which one they’re using right now, but what the interest is in identifying users who are similar or have similar types of behaviors to the users they give recommendations to. So that may be things like, they tend to, people who watch certain types of movies tend to have certain demographic features, or live in certain areas or be certain ages. There’s a lot of different types of inferences that they make about individuals and about their behavior and their patterns in order to decide on what movies to recommend and how to rank them.

How does the recommendation engine function?

Lisa Singh: It’s a combination of things. So one is what have you indicated your interests are in. So what movies have you liked, what do your reviews look like for different types of movies, and what types of genres do they fall into. So they’re trying to understand first what you seem to like. Then they identify individuals who seem to have similar preferences and identify other movies that you haven’t identified or shows that you haven’t identified, but those other users with very similar profiles have identified as being ones that they like and then they recommend those ones to you. In order to decide on the ranking, not only will use the features about that movie but then they’ll also start to use the features of the users that have similar tastes as you do. So they might identify the fact that women have similar taste. They might identify the fact that they think that you’re in a particular age group. Those are all different types of features that they would do at that stage.

Do you use the Netflix Recommendation System?

Matthew Tinkcom: I see it, but I don’t find myself using it. And I think it’s because of this funny thing. When I’m done watching a particular show or movie, I’m usually not hungry for that same kind of show or movie, I want something different. And I find that the recommendations usually , they want me to watch something or they encourage me to watch something nearly identical. So if it’s patterning my behavior and I understand why it does this, it assumes that I’m only interested in that kind of show because I’ve just watched 20 episodes of it.  

Is this technology bad for user privacy?

Lisa Singh: In general, as long as the data you choose to share is all they’re using, and they’re using it only for things like recommendations or things to help with their core business, that you’ve chosen to share it with them so I think it’s okay.  It’s when they start sharing it with other people or when they use it for other purposes than you think they’re using it for or things you don’t know about, then it becomes a privacy issue and that’s something we should all be concerned about. Our data is our currency in today’s market and we need to make sure that we value our data and we don’t let companies just use it any way they want without thinking about the consequences about they way they use it.