Ideas for Recommendations

Information about user

• solution times
• subjective flow-feeling (reported perceived difficulty)
• click on "Give me easier task" button
• click on hint button

Recommendation criteria (for scoring formula)

• time since the last solution
• flow estimate and/or student-task difficulty estimate
• task efficiency (how much the student will learn by solving the task)

Recommendation models

• Elo variations
• Simple neural network model
• Restricted Boltzmann machine (bipartite graph of neurons: one side for student, one side for task)
• Evolution algorithms for improving constants/parameters estimation (or just multi-armed bandit)