The Role of Artificial Intelligence in Educating Novice Programmers

Programming is an inherently difficult skill to acquire and develop. Those who attempt to learn programming may be easily discouraged. The current landscape for computer science education does not address the needs of every novice programmer. Literature reports a discrepancy between student misconceptions and instructors’ perceptions of those misconceptions. Those who can afford a one-on-one human tutor perform on average two standard deviations better than those who learn via conventional methods, suggesting there is a need for a comparable, cheaper replacement. As a result, a number of intelligent tutoring systems have been developed for the purpose of teaching introductory programming concepts and replicating the benefits of one-on-one human tutoring. In this thesis, we analyze and discuss the literature pertaining to student misconceptions, selecting five fundamental misconception categories for introductory programming to demonstrate the effectiveness of existing intelligent tutoring systems. The features of existing intelligent tutoring systems are discussed and analyzed with respect to their effectiveness in addressing student misconceptions. Finally, we highlight the current gap in research on intelligent tutoring systems, hypothesizing the architecture and features of an ideal intelligent tutoring system for introductory programming.