The Momentum-Exploration recommender system can be thought of as the inversion of a centrally planned curriculum.
The name Momentum-Exploration can be understood based on an analogy with the explore vs exploit problem: the recommendations made by the system would exist along a spectrum from appealing to the interests the system knows the learner already has to giving recommendations from new unexplored territory.
In exploitation mode, the system would not prescribe a diet of information to take you from A to B like the centrally planned curricula of the traditional education system, rather it would function like YouTube's recommendation algorithm - it would be optimized solely to help you find the next best article, book or video. It would not bother to search much beyond the next article, book or video because the system would be built with the understanding that the learner's prior knowledge and interests are constantly dynamically updating so only the learner's interactions with the next piece of content can define the next best step.
However, the algorithm would differ from YouTube content recommender because it would not only have a deep understanding of the learner's prior behaviour and history of interaction with the system, but also their prior knowledge.
By analysing the learner's database of spaced repetition prompts, including repetition histories, retrieval probabilities and lapses, the system could have an understanding of the current state of the learner's model of reality.
Likewise, by processing the learner's journal, notes and articles, the system could understand