Perusall takes a very careful approach to the use of AI. Models are never trained on your data. Our guiding principles are that AI in Perusall should be pedagogically grounded, safe, private, and optional:
- AI in Perusall is always pedagogically grounded. We never use AI just for the sake of using AI. Instead, we use AI when our educator-led team believes it provides serious pedagogical value for your course.
- It's safe to use Perusall AI in your course. Data security is paramount and we have extensive protections in place to ensure that your data is handled in accordance with our strict Privacy Policy.
- Your privacy is preserved. Student and instructor data is never sent to model providers; all models are run inside a closed system (our private cloud computing environment).
- Use of Perusall AI is always optional. Instructors always have full control over whether and when to use AI in their courses.
Below are specifics about which features use AI, and how:
- This is a supervised machine learning approach to automatically grading comments and questions that students post, for thoughtfulness and effort. Perusall suggests quality scores for each comment and question posted by students, which instructors can optionally integrate into the student’s score for the assignment. Instructors always have full control over whether and how these scores are used, and can override them at any time.
- Instructors can add checkpoints to audio, video, and text-based standard assignments using generative AI to suggest a discussion prompt or generate quizzes. Instructors have the ability to preview, edit, and accept or discard the suggested AI material before incorporating it into an assignment.
- For each assignment, instructors have access to a “confusion report” that automatically summarizes the questions that students are asking about the assigned content. Student questions are grouped into related topic areas, with specific questions asked by students called out for you as exemplifying each topic. This allows instructors—particularly in large courses, where students write more questions than an instructor would have time to read—to use student questions to effectively plan out their class time and make connections between the questions students asked in the assigned reading (or video, podcast, etc) and what happens in class.
Related to