Skill builder problem sets have the following features:
Questions are based on one specific skill, a question can have multiple skill taggings.
Students must answer three questions correct in a row to complete the assignment;
If a student uses the tutoring (“Hint” or “Break this Problem Into Steps”), the question will be marked incorrect;
Students will know immediately if they answered the question correctly;
If a student is unable to figure out the problem on his or her own, the last hint will give the student the answer;
Currently, this feature is only available for math problem sets.
Use case
Adaptive Skill Mastery through Reinforcement Learning in Intelligent Tutoring Systems
| Example observations | student_id (one-hot or embedding) current_skill_mastery (vector for 100 skills, cumulative proportion of correct answers) current_skill_id problem_id opportunity (number of prior attempts on this skill) correct_prev (0/1) |
| Example actions | action can be index of the next problem that was actually presented to the student by the platform There are approximately 13,000 distinct problem statements; the action space can be reduced to: selecting a skill (100 options), then sampling a question with increasing difficulty, or selecting a (skill, difficulty) pair (~300 possible actions) |
| Example reward | +1 if the student answers correctly on the first attempt, 0 otherwise |