Member-only story
A recommender system starts with some sort of data about every user that it can use to figure out that user’s individual tastes and interests. Then it can merge its data about X (a specific user) with the collective behavior of OTHERS like X to recommend stuff X might like.
But where does that data about X’s unique interest come from?
There are two ways, and our intention is to discuss them in brief.
EXPLICIT FEEDBACK:
One way to understand users or customers is through explicit feedback. For example, asking users to rate an online course on a scale of one to five stars, or rating content they see with a like or a thumbs up or a thumbs down.
In these cases, we are explicitly asking the users — ‘do you like this thing you’re looking at?’ And we use that data to build up a profile of that user’s interests.
The problem with explicit ratings or feedback is that it requires extra work from your users. Not everyone can be bothered to leave a rating on everything they see, so this data tends to be very sparse. And when the data is too sparse, it leads to low-quality recommendations.
Another problem with explicit ratings is that everyone has different standards, so the meaning of a four-star review…
