Abstract
More than one third of computer science students either switch disciplines or fail the first programming class due to test/performance anxiety or learning difficulties. Personalized feedback can potentially improve both learning outcomes and retention for novice learners of computer programming. Online learning management and assessment systems have the power to allow self-paced, individualized and multi-layered learning that can be easily combined with time tested learning strategies such as pair programming, game based learning or visual programming to positively impact learning outcomes and support free-choice learning. In this paper, we introduce an online automated tutoring and assessment system, called MindReader, for introductory programming in C++. We discuss how MindReader generates personalized feedback fully automatically and prioritizes semantic error messages to avoid overwhelming the learner with a large number of signs of "failures." We also discuss how MindReader can be used as a flexible teaching tool for introductory programming classes, and how its smart tutoring and assessment systems can improve learning.