This seminar-oriented course will look at the emerging science of designing large scale learning systems, particularly those that rely on peer interactions. Classrooms based on peer interactions (such as peer assessment, discussion, and tutoring) can dramatically improve learning -- for example, they have been shown to halve the failure rates in introductory CS classes. How can we design computer systems that enable and enhance these benefits at large scales, with thousands of participating students? This course will introduce you to the theory of peer learning, to design patterns that embody this theory, and case studies that demonstrate how these patterns are implemented in applications used by tens of thousands of students. As case studies, we will look both at commercial products (Coursera, DuoLingo...), as well as research projects we have developed that are used by 10K+ users.
A few of the topics we cover:
- How do you scale peer assessment to more than 100,000 students in 100+ MOOCs?
- How do you use simple machine learning techniques to improve student work?
- How do you use massive scale to provide fast (sub 10minute) feedback at any time of day?
- How do you leverage diversity in global classrooms to spark new thought?
- What is the role of teachers in a large-scale peer-assisted classroom?
- What are the ethical questions around large-scale learning?
This class involves a mini-project.
Feedback from students:
- "I really enjoy your class and it leads to some of the best and most interesting discussions I have had at CMU. This class is forcing me to see various sides of the argument and really testing my own belief and idealogical system, which is really good"
- "I want to say that I’ve enjoyed your class greatly so far this semester. It has very often felt like the necessary counterpart to my Design classes — being based in reading, thinking and making instead of making and ‘crit-ing’. Your reference to Design and Art examples, specifically, has been of consistent interest."