Adviser for algorithms classes

Computer and data science professionals are in higher demand than ever before. Students taking computer science classes has increased immensely. In the United States in 2015, there were nearly 10 times as many advertised computing jobs as there were students graduating with a bachelor’s degree in computer science. At the same time, US institutions are struggling to retain or hire CS faculty, making it hard to directly address the issue.

This project will seek answers to a number of questions: Can students understand the comments? Do instructors want to keep using it? Does it cause better student learning? Do students believe they receive worthwhile feedback? Do students feel more engaged in the course?

Can we develop AI technology to assist instructors in their future work? One aspect of instructor’s jobs is providing feedback to students’ explanations in the classroom and on homework. Research in cognitive science strongly supports the premise that students can more effectively learn concepts if they explain answers to cognitively demanding “deep” questions in their own words, as concluded by a U.S. Department of Education blue-ribbon IES Report (Pashler et al., 2007).

When a teacher decides not to send any of the three examples, that datum is also stored as part of the training set). As course-staff decide which messages make sense, the machine learned component can get better and better.

Exhibit 4. The ADVISER proposed interface showing three different suggestions for four out of five students. For the student named "Sachi," the instructor has changed the diagnosis and written her own comment.

In this project, Dr. George Heineman and Dr. Neil Heffernan will develop Smart-Reply tools for teaching assistants. These replies will allow TAs to correspond with students and give frequent feedback in a time-efficient manner.

Increasing the efficiency of feedback will allow TAs to be more effective and allow them to help more students!