Chapter 19 Authoring Problem-Solving Tutors: a Comparison between ASTUS and CTAT Luc Paquette, Jean-François Lebeau and André Mayers Computer Science Department, Université de Sherbrooke, 2500 boulevard de l’Université, Sherbrooke, Québec, J1K 2R1, Canada { luc.paquette ; andre.mayers@USherbrooke.ca} Abstract. ASTUS is an Intelligent Tutoring System (ITS) framework for prob- lem-solving domains. In this chapter we present a study we performed to evaluate the strengths and weaknesses of ASTUS compared to the well-known Cognitive Tutor Authoring Tools (CTAT) framework. To challenge their capacity to handle a comprehensive model of a well-defined task, we built a multi-column subtrac- tion tutor (model and interface) with each framework. We incorporated into the model various pedagogically relevant procedural errors taken from the literature, to see how each framework deals with complex situations where remedial help may be needed. We successfully encoded the model with both frameworks and found situations in which we consider ASTUS to surpass CTAT. Examples of these include: ambiguous steps, errors with multiple (possibly correct) steps, com- posite errors, and off-path steps. Selected scenarios in the multi-column subtrac- tion domain are presented to illustrate that ASTUS can show a more sophisticated behavior in these situations. ASTUS achieves this by relying on an examinable hierarchical knowledge representation system and a domain-independent MVC- based approach to build the tutors’ interface. 19.1 Introduction Intelligent Tutoring Systems (ITS) that support a learning-by-doing pedagogical strategy are usually developed in line with one of three established approaches: model-tracing tutors (Anderson and Pelletier 1991), constraint-based tutors (Mi- trovic et al. 2003) and example-tracing tutors (Aleven et al. in press; Razzaq et al. 2009). All of these fit VanLehn’s tutoring framework (VanLehn 2006), in which a model of a task domain is used to evaluate each of the learner’s steps (which are themselves driven by mental inferences) as correct or incorrect. Model-tracing tu- tors such as Cognitive Tutors (Anderson 1995) and Andes (VanLehn 2005) have