Automated advice-giving strategies for scientific inquiry

Abstract

We describe a prototype advisor for students using Belvedere, an environment for conducting discussions about scientific controversies. The advisor has two strategic components, syntactic and consistency-based. Syntactic strategies are based on structural and categorical patterns in argument representations constructed by the students, and suggest ways in which students can continue their inquiry. Consistency-based strategies check student-made links between pairs of statements against the pairwise relations specified between corresponding units in a knowledge base constructed by a teacher or expert, and identify information that may challenge or corroborate relationships proposed by the students.