Student misconceptions of reinforcement learning

[Sverrir Thorgeirsson]

Dear all,

My colleagues and I are building a validated diagnostic test to assess students' working knowledge of reinforcement learning. Standardized tests of this type can be important for instructors, but they require input from many experts from around the world.

If you teach RL and know what kind of RL-related misconceptions students typically experience, we would love to hear from you. We will send three short surveys using a process called the Delphi method. In the first round, we ask you to name topics and misconceptions related to RL. In the next two rounds, we ask you to anonymously rank the responses from the other experts. At the moment, twelve experts have volunteered their help; we are hoping to find three more.

In return for your help, we will acknowledge your contribution in all project outputs, including publications.

To participate, please email me (sverrir.th...@inf.ethz.ch) or register here:

https://forms.gle/pUP1xKQjEehoMhbn8

Thank you very much!
Sverrir

[Sverrir Thorgeirsson]

Dear all,

Our diagnostic test on introductory RL is now ready. I am very thankful to everyone who participated in the expert panel; the response rate was very high.

We are now looking to deploy our test so that we can validate the instrument and gain stronger insights into common student misconceptions on RL. If you are teaching RL and would like to administer the test in your course or with your students, we would be thrilled to work with you. The test takes 30 minutes for students to complete, has 13 multiple-choice questions, and can be taken online. The questions cover topics like TD learning, policy gradients, and value function approximation, and can be administered either as a pretest (to learn how much students know before instruction begins) or as a posttest.

Thank you very much,
Sverrir