We use recommender system all the time. A website will recommend something to you based on what you’ve watched, listened to, bought or who you’ve friended on Facebook. These systems attempt to predict your preferences based on past interactions.
The systems range from simple statistical approaches like Amazon’s people who bought X also bought Y links, to complex Artificial Intelligence-based approaches that drive feed ranking on sites like Facebook.
Julian McAuley, UC San Diego Computer Science and Engineering, explores the modeling techniques behind personalized recommendation technology on the web and the different systems that we encounter.
He reminds us that we often find these recommendations a bit “creepy,” but that actually the recommender system has no human intelligence; it’s really a simple statistical process. They focus on short-term predictions but could they be adapted to make long-term predictions or estimate more subjective qualities? Would that be bad?