Chapter 13: Continuous student modeling for programming in the classroom: challenges, methods, and evaluation
Restricted access

Student modeling that can accurately predict learning outcomes is often crucial for developing effective, personalized Intelligent Tutoring Systems. To build effective student models, we need to determine what things to track (skills) and how to track them (modeling). In this chapter, we focus on the domain of programming, where we may not have labels for these skills, and it is challenging to recognize discrete steps and to assess associated student performance. We present the novel techniques our research group has derived for addressing these challenges by considering data-driven skill discovery, temporal modeling, and semi-supervised methods. Our findings suggest that feature-based models can be good choices to handle large solution spaces, but taking into account both structure (e.g. abstract syntax trees), and time (e.g. temporal information), using neural network and pattern mining mechanisms, can produce more effective models. Finally, we present a multi-criterion evaluation mechanism to understand the impact of adaptive support systems that leverage data-driven student models for programming.

You are not authenticated to view the full text of this chapter or article.

Access options

Get access to the full article by using one of the access options below.

Other access options

Redeem Token

Institutional Login

Log in with Open Athens, Shibboleth, or your institutional credentials

Login via Institutional Access

Personal login

Log in with your Elgar Online account

Login with your Elgar account