Personalized Thinking Model for AI Education with 75.48% F1 Score
A new educational initiative called the Personalized Thinking Model (PTM) has been introduced to enhance artificial intelligence applications in learning environments. PTM categorizes insights from student journals into five distinct levels: behavioral instances, patterns, cognitive routines, metacognitive tendencies, and self-values. Rooted in Marzano's Revised Taxonomy, this model aims to form a cognitive profile of individual learners. Utilizing advanced techniques including large language model inference and consensus clustering, a seven-week study with 40 participants resulted in an initial F1 score of 74.57%, which improved to 75.48% following a refinement process emphasizing atomic information.
Key facts
- PTM is a hierarchical and interpretable learner representation for AI-supported education.
- It organizes evidence from learner journals into five layers: behavioral instances, behavioral patterns, cognitive routines, metacognitive tendencies, and self-system values.
- PTM is grounded in Marzano's New Taxonomy of Educational Objectives.
- The model aims to clone a learner's thinking model and build a cognitive twin.
- Construction pipeline uses Gemini 2.5 Pro, sentence embeddings, dimensionality reduction, and consensus clustering.
- Evaluation involved 40 participants in a seven-week study.
- Automatic evaluation using atomic information point matching yielded F1 scores of 74.57% before HITL and 75.48% after HITL refinement.
Entities
—