To Tutor or Not to Tutor: That is the Question

Leena Razzaq; Neil Heffernan

Artificial Intelligence in Education (AIED) -- 2009

Keywords

Intelligent tutoring system; Tutored problem solving; Scaffolding; Feedback; Time on task

Research Brief

This study compared tutored problem solving, or working through problems step-by step, with worked solutions while controlling for time on task. Using a counterbalanced design to control for order effects, the authors randomly assigned 88 8th students to two ASSISTments assignments, Symbolization and Slope & Intercept, with either tutored problem solving (see Figure 1) or worked solutions (see Figure 2) provided as feedback, as shown in Table 1. The authors hypothesized that highly proficient students would benefit more from worked examples while less proficient students would benefit more from tutored problem solving.

Of 83 students who completed both assignments, normalized gain scores from pre- to post-test showed significant learning gains in both assignments, p < 0.01. Students completed significantly more problems with worked solutions in both problem sets. A median split on prior MCAS performance was then used to establish groups of high and low proficiency students. Repeated measures analyses revealed an interaction effect: when controlling for time on task, proficient students exhibited significantly more learning from worked solutions than tutored problem solving while less proficient students exhibited slightly more learning from tutored problem solving. Less proficient students also spent significantly less time on problems with worked solution feedback.