Recommender Systems for Online Education
I've read a lot recently about recommender systems, which have been defined as "personalization tools that help users to find interesting information and services in complex online shops". Amazon is one of the best known sites that uses them, and they were smart enough to provide a bypass mechanism so that the present you bought on a topic only of interest to the recipient doesn't trigger new recommendations to you.
In e-learning, pretty much everything is online, providing massive amounts of data that can be monitored and analyzed. Yet I don't know of online courses that recommend "since you spent a long time reading that, you might also enjoy this" or even "here are some other courses you might find useful". Is it that the data is different, or harder to mine, than Netflix data; that students don't want or need recommendations; or that education is not a commercial venture (at least not in the same way as Amazon or Netflix) and the goals of recommender systems are not applicable here?
Comments
There have been a number of recommendation tools for learning (and specifically online learning), but mostly these are based on models of material mastery (deep user models) rather than data mining approaches such as collaborative filtering.
I think the biggest reasons why are:
(1) It is rarely the case in education that there are enough options at any point in an educational sequence. In theory, it would be nice to have 20 different ways to learn a unit a electromagnetism, and to pick the right ones based on your experience and the experiences of others, but this requires a lot of course modules.
(2) Education is a harder problem because of sequential dependencies. It is easy to recommend a book to someone without worrying (except in rare cases) that this book only makes sense before a certain other one, or after a certain other one. Education is more cumulative.
Posted by: Joseph A. Konstan | October 6, 2009 01:03 PM