A new system adapts to help people learn physical skills

An online personalization algorithm adjusts tutoring goals based on a learner's confidence, workload, and performance.

Researchers developed an online personalization algorithm for psychomotor learning. This system continuously changes tutoring priorities to match the changing needs of a learner.

The study used a 2D quadrotor landing task with 40 human subjects. The algorithm uses an actor-critic-inspired method to update reward weights. These weights come from recent measures of performance, workload, and self-confidence.

Psychomotor learning involves several objectives. These include driving skill progression, calibrating workload per the Yerkes-Dodson Law, and calibrating self-confidence to performance. A multi-input, multi-output cognitive control framework synthesizes tutoring policies through model-based planning.

The algorithm showed different patterns for different learners. Learners whose weights shifted more substantially often reached higher final performance. Those who consistently emphasized workload calibration achieved the strongest outcomes.

Existing psychomotor intelligent tutoring systems (ITSs) often rely on models built before instruction. It remains open how to use real-time assessments of cognitive state for finer-grained personalization. Balancing multiple tutoring objectives over time also remains open.

Key facts

  • The study involved 40 human subjects using a 2D quadrotor landing task.
  • Learners who emphasized workload calibration achieved the strongest outcomes.
  • The algorithm uses an actor-critic-inspired method to update reward weights.

All sections

Edition