This study aims to improve the reliability of a new AI collaborative scoring system used to assess the quality of students’ written arguments. The system draws on the Rational Force Model and focuses on classifying the functional relation of each proposition in terms of support, opposition, acceptability, and relevance.