Prof. Kobi Gal

We study content recommendation in an online learning environment for mathematics (N=77, 4^th-5^th grade student). We compare an expert teacher's recommendation to that of a neural network algorithm, implementing collaborative filtering ranking. We do so using a double-blind randomized controlled experiment. We find that when the difficulty of the teacher's sequence of recommendation was overall increasing, the teacher was superior to the algorithm regarding students' performance. Taken together, our findings indicate on how the algorithm and the expert teacher can benefit from each other.