QUICK-Comments is an AI-powered program that will suggest personalized immediate feedback comments to teachers that they can send to their students. Current math textbooks ask students to explain their thinking 3 to 4 times on each night's homework, but teachers often don't have time to give detailed feedback on these problems. ASSISTments has traditionally graded multiple choice or problems with numerical or expression answers, but cannot grade the short explanations that students write for open-response questions.
That's where QUICK-Comments comes in.
An example of Google Smart Reply's suggested responses
QUICK-Comments will rely on machine learning and natural language processing (NLP) techniques to suggest responses for teachers to send to their students. This will work in a manner similar to Google Smart Reply, which provides Gmail users with three suggested responses they can send without having to craft their own message.
The first step in creating the tool will be to collect actual teachers’ comments that we can use as a training set for machine learning. We have already received grant funding to pay a small cohort of teachers to read their students' open-ended responses and write comments. Once this has occurred, the team will apply machine learning in order to anticipate the correct response for future student writing.
A second group of teachers will be able to choose from three comments; some choices will be teacher-created, while others will be our machine learned sequence-to-sequence generated comments. We call this process the Active Machine Learning Paradigm, as each night we will refit our machine learned models to try to use all the data we gather each day. We will not only learn which suggestions teachers pick, but we can learn a lot each time a teacher decides not to send any of our three suggested comments.
The proposed QUICK-Comments interface shows three different suggested comments for each student listed.
Cristina Heffernan recently presented a keynote address at the Korean Women in Mathematical Sciences conference about Quick Comments!
Daily communication has become easier and quicker with the help of smart technology. Suggested replies and text messages allow us to easily click and send rather than spending time writing. In this presentation, you will learn about how artificial intelligence is advancing the mathematics education field and how teachers play a role in those enhancements. This is the story of how we created QUICK-Comments a feature of ASSISTments.
New Accepted paper: Baral, S., Botelho, A., Erickson, J.A., Benachamardi, P., & Heffernan, N. (2021). Improving Automated Scoring of Student Open Responses in Mathematics. Educational Data Mining. EDM 2021 PDF https://educationaldatamining.org/EDM2021/virtual/static/pdf/EDM21_paper_188.pdf Camera Ready Copy. Best Full Paper Nominee. Presentation is here
Zhang, M., Baral, S., Heffernan, N. & Lan, A. (2022). Automatic Short Math Answer Grading via In-context Meta-learning. Accepted to EDM2022. Submitted version
Baral, S., Seetharaman, K., Botelho, A., Wang, A., Heineman, G.,& Heffernan, N. (2022) Enhancing auto-scoring of student open-responses in the presence of mathematical terms and expressions. AIED2022. Submitted version - Shorter Final version. Video