During section:
The goal of this section is for you to try to use a GPT-Edu (not AI sandbox) to solve a homework assignment from a course that you’ve never taken.
Your TF will organize you randomly into pairs; if there’s an odd number then the TF will join with a student to make one of the pairs. One student from each pair will give their selected assignment question to a different pair.
Working in pairs, you will have 15 minutes using GPT-Edu to complete the assignment that you received. (Be sure that neither you nor your partner have seen the assignment before.)
Then each person who gave an assignment to another pair will leave their current group and go over to the group that tried to solve the assignment. (“follow your assignment”). Take about 5 minutes to critique the answer that they obtained, assign a grade, and attempt to improve the answer using the LLM. Stay at that location (this is your new pair for the next exercise).
Now, repeat the whole process with the assignment from the other student in the pair (the one who didn’t move). They will give their assignment to a different pair, allow 15 minutes for that pair to try to solve it, and then move over to assess that solution.
At this point, everyone will have attempted to solve 2 different assignments, and each assignment will have been attempted by one pair using the LLM. Your TF will collect all the feedback about these assignments and summarize it on the board. Then they will lead a discussion about the results: which attempts were successful, and why? Was there any pattern that you can glean from the successful vs. unsuccessful assignments? Did it make a difference if you had some content knowledge related to the assignment? What did you learn from this activity?