Skip to content

How Could Teaching Assistants Cooperate with AI Tutors?

In this session we invited one CS graduate who helps develop an AI tutor system and one TA who deployed an AI tutor.

Icebreaker: classic self-introduction

I am a TA for course XX, then choose one of three questions to answer:

  1. From a technical standpoint, which work is most likely to be taken over by AI? (e.g. some OOP code tutoring)
  2. From a technical standpoint, which work is least likely to be done by AI? (e.g. final grade aggregation, or "no step should be handed to AI", code correctness)
  3. From personal preference, which work would you most like AI to take over? (e.g. handling unstructured data, tutoring, personalized comments, retrieving similar records in the course group, copying document answers, screening hundreds of lab reports, etc.)

A point worth pondering: which TA work is "technically feasible but perhaps shouldn't be handed to AI," and which is "something you'd love to offload to AI but can't yet" — these two sets don't coincide.

Sharing

Niu Songjie: deploying an AI tutor in a course

Using Computer Culture Basics as an example, introducing the Moobius AI tutor "parallel classroom", Stable Diffusion AI drawing, and AI-computing popularization micro-seminars.

Feng Zhuoer: how an AI tutor is developed

A graduate student involved in developing an AI tutor system sharing the process from requirements to implementation.

Discussion

In a "Tools in TA's work" discussion, a series of efficiency-improving TA-work questions was raised, but the "solutions" weren't explored enough at the time. This session revisits that and does a simple reflection.