If you start with an algorithm that accepts any type of user behavior (is multimodal) then using user-dislikes is easy. We can easily do this with the Universa

If you start with an algorithm that accepts any type of user behavior (is multimodal) then using user-dislikes is easy. We can easily do this with the Universal Recommender as shown in this article (reprinted from the IBM dev blog) https://actionml.com/blog/making_dislikes_predict_likes https://actionml.com/blog/making_dislikes_predict_likes