Peer Feedback

I founded Peer Feedback in December 2015 to handle peer review in large online courses, where a handful of TAs can’t read every assignment closely. I built the platform from the ground up, integrated it with Canvas LMS so courses could import rosters and assignments, and added research-based collaborative learning features. I ran it until I sold the company in 2025.

Georgia Tech was the biggest user. According to Georgia Tech’s FY20 Technology Fee request, filed by Dr. David Joyner, “over 323,000 peer reviews have been generated by over 11,000 students.” That request asked for $118,000 to fund “Peer Feedback Enterprise.”

Scoring the quality of every review

Peer review only helps if the reviews are good, and with that many of them nobody can check by hand. I hired an ML engineer and led the effort to build a deep-learning model that scores the quality of every written peer review. It’s a three-class classifier trained on years of meta-feedback, where students rated how useful the reviews they received were. We rolled it out to instructors and TAs first, so they could spot weak reviews, and then to students, who see the score while they’re still writing.

Research and talks

• Joyner, D. A., Perez Irizarry, G., Eicher, B., & Lytle, N. (2024). Intelligent Feedback and Evaluation for the Assessment and Improvement of Student Peer Reviews. IEEE Digital Education & MOOCs Conference.
• Co-presented “Leveraging Peer Review to Support and Enhance Learning at Scale” with Dr. David Joyner at Georgia Tech’s Teaching with Technology Spotlight, April 16, 2018.