An intelligent tutoring system that generates a natural language dialogue using dynamic multi-level planning

Area of Science:

• Physiology Education
• Artificial Intelligence in Education
• Natural Language Processing

Background:

• Students require practice in qualitative causal reasoning for effective knowledge internalization and application.
• Learning is enhanced when students articulate their understanding in words.
• Previous research highlighted the need for interactive learning environments in physiology.

Purpose of the Study:

• To develop an intelligent tutoring system (ITS) for physiology.
• To enable the ITS to conduct natural language dialogues with students solving physiology problems.
• To facilitate qualitative causal reasoning practice through interactive dialogue.

Main Methods:

• Analysis of 75 hours of physiology tutoring sessions informed system design.
• The system simulates cardiovascular variable responses to blood pressure perturbations.
• It employs a natural language understanding component using finite-state machines and lexical functional grammar for text generation.

Main Results:

• One hour of system use yielded significant learning gains.
• The ITS improved problem-solving abilities more than traditional textual material.
• Student feedback indicated positive reception and perceived learning benefits.

Conclusions:

• The CIRCSIM-Tutor system demonstrates the efficacy of ITS in natural language dialogue.
• Current language technology is sufficient for implementing effective AI-driven tutoring.
• The system is integrated into the first-year physiology curriculum at Rush Medical College.