We present CritiqueKit, a mixed-initiative machine-learning system that helps students give better feedback to peers by reusing prior feedback, reducing it to be useful in a general context, and retraining the system about what is useful in real time. CritiqueKit exploits the fact that novices often make similar errors, leading reviewers to reuse the same feedback on many different submissions. It takes advantage of all prior feedback, and classifies feedback as the reviewer types it. CritiqueKit continually updates the corpus of feedback with new comments that are added, and it guides reviewers to improve their feedback, and thus the entire corpus, over time.
No takes yet. Share an insight, caveat, or question.
Fraser et al. (2017) studied this question.