Skip to content

Orientation: Common Traps and Efficiency Tips for TA Work

Format:

A 10-minute brief lecture + a 50-minute panel discussion (four experienced TAs from different course types).

The 10-minute lecture centers on five common traps new TAs may face.

The 50-minute panel discusses pre-arranged questions; audience can submit questions via a questionnaire at any time for the panelists to discuss.

Finally, the audience is introduced to the plan for the rest of the semester's TA training.

A. The 10-minute lecture

A1. The role of TA training

Have you thought about why you attend TA training — is it just to take up your precious time? From the school's perspective, TA training aims to "improve undergraduate teaching quality and promote the comprehensive development of graduate students." That's not wrong, but it's a bit far from you.

A good TA experience is rewarding. You take on a reasonable workload and gain a sense of achievement from positive feedback. You might feel: although my research isn't very meaningful, my TA work is. Then there's exploring infinite possibilities. In research you can't always choose your own topic. But when you design an assignment in a course, you can explore many interesting possibilities, like telling a story in the problem description.

TA work can also keep you from being lonely. Research is a lonely journey, but TA work helps you make more friends — students in your course, and other TAs. Didn't we create a TA group? It's not just for announcements; whatever problems or confusions you have in TA work, you can confidently ask in the group, hoping to reduce your loneliness.

But TA experience is not necessarily positive. I know students who had bad experiences as TAs — feeling a heavy workload and exhaustion, making mistakes that harm students while leaving psychological shadows on themselves, or feeling a sense of emptiness about TA work.

In my view, the most important goal of TA training is not to make TAs do things for teachers or students. Rather, it's to make our own TA experience happier and more meaningful. So the first session's theme is: efficiency tips and common traps for TA work, hoping you won't feel exhausted, and won't leave yourself psychological shadows from common mistakes.

A2. Five common traps for new TAs

  • Working by feeling/passion without clear division of labor, planning, and basis.
  • Subconsciously seeing yourself as a student, without completing the identity change from student to TA.
  • Assuming students are like yourself, or treating yourself/students as machines.
  • Believing TAs should reply to students as promptly as possible in any case.
  • Believing TAs should try to satisfy every student's every request.

(Note: these five traps come from interviews with multiple CS graduate-student TAs.)

Trap 1: working by feeling/passion without clear basis, division of labor, and planning.

Consequence: TA work becomes chaotic, time is tight.

Corresponding efficiency tip: have a basis, good division of labor, and planning.

Basis: written materials — the detailed syllabus/course announcements — are the basis for communicating with students. Also teaching-team consensus and conventions, which may not be written. As a new TA, get familiar with these materials, including late-submission and plagiarism policies. Ask the teacher or previous TAs for materials or chat over a meal.

Division of labor: between TAs, and between TAs and the teacher. Clarify what you need to do / don't need to do; what you can / can't do. The teacher or head TA should ensure the workload is fairly distributed.

Planning: when TA work needs to be done. Most TA work has no deadline; we need to schedule it ourselves, ideally reserving time at the start of the semester. For example, a fixed weekly slot for TA work. Or figure out when assignments are due, so you know which weeks you need to work. If you use an electronic calendar, add important assignment deadlines, like you would meetings.

Trap 2: not completing the identity change from student to TA.

Consequence: misunderstandings in teacher-student communication, harming the course and personal image.

Being a student in class and being a TA are different; don't bring student habits over. First, you can't speak carelessly. Every sentence represents the teacher and the teaching team. Even in private messages, assume it might be screenshotted and posted to Moments.

As students we could complain about TAs and courses. But as a TA, remember: you can't complain about students or speak ill of them; you should only give constructive learning advice. If you want to vent, vent to ChatGPT.

As a student, you could be wrong while helping a classmate; but as a TA, you shouldn't. Helping a roommate, you don't have to be right. As a TA, even saying "I don't know" beats giving ambiguous or wrong answers.

Some TAs have reform ideas about the course; note that ideas from your student days may not suit the actual teaching. Proceed carefully, understand the real situation, and don't treat students as guinea pigs.

Trap 3_1: assuming students are like yourself

Consequence: misunderstandings in assignment design and teacher-student communication.

Before, you may only have communicated with familiar people; as a TA you communicate with unfamiliar ones. We used to think the people around us were everyone, and that our communication style applied to all. But as a TA you meet many people you haven't before, so you can't communicate by past experience/assumptions. And you can't assume others think like you. We need a professional communication style, like a doctor with a patient. Not that students are sick — rather, learn the doctor's communication style: clear, and humane.

For example, avoid overly in-group speech. Don't tell a student "Read The Fxxking Manual." A student will think "who are you?" and feel offended, not think it's cool.

When estimating assignment workload, start from an ordinary/newcomer student's perspective, not the TA's expertise.

Trap 3_2: treating yourself/students as machines

Neither students nor TAs are machines: humans aren't grass or trees; who has no feelings? We make mistakes, get emotional, forget. So treat people as people, not machines. Although the interface for chatting with students in WeChat is like chatting with ChatGPT, it's different underneath. A student isn't an LLM; if we hurt their emotions, they really lose their dopamine. So when you do something wrong, apologize quickly — don't be stingy with apologies; apologize when you should. (If a careful assessment says you should stick to principle, then do.)

Also because people aren't machines, we need to reduce ambiguity in communication. For example, if we want lab reports to be shorter for faster grading, and we say in the group "please write reports shorter," students still write very long reports because they're not sure how short is appropriate. So don't make students guess our intent. Provide a word limit (deduct beyond it), or even a template.

Course announcements should be humane, not bring negative emotion, and not treat students as enemies. When making rules, explain the intent, or students feel puzzled or uneasy. Some rules may have special reasons from the teaching team, but students don't know, so they're confused.

Also, students may be reluctant to seek feedback from TAs. When announcing and assigning, remind them to ask questions through a certain channel, guiding them to ask. Even if those channels exist without reminder, students won't ask if we don't remind.

Important messages should be repeated. Repeated. Repeated. Repeated.

Trap 4: TAs need to satisfy every student's every request

We can only try to satisfy most students' reasonable requests.

Don't rashly agree to student requests — e.g. a lone request to resubmit homework. Ask: does every student have this opportunity? Even if it seems reasonable, if only some have it, no. If you allow this student to resubmit, ensure: post an announcement telling all students they have a similar chance. Also the deduction amount, or deduction curve, is best published.

For example, don't delay a deadline close to the deadline. When many request a delay, use a questionnaire to decide — each student votes delay or not, and the majority wins.

Trap 5: TAs need to reply as promptly as possible in any case

Consequence: TA work consumes significantly more time, with a much higher chance of error.

There are different cases: content questions, and course-management questions (e.g. resubmission).

CS students know the memory-hierarchy concept — memory, disk, different performance levels, different layers to organize data.

Actually, student tutoring has a similar hierarchy. Search engines and LLMs have excellent availability, usable anytime, but their quality should be worst. Other channels — ask a roommate, the tutoring center, the TA — have worse availability but higher quality.

Quality here means truly solving their problem, giving correct answers, and helping them learn more in the process.

If being a good TA only required prompt replies, ChatGPT and the tutoring center would do better than us. So being a good TA now should value tutoring quality: I can't give prompt replies 24/7. But every exchange with a TA should fundamentally solve your problem and help you learn/improve.

Use email (making clear TAs have no obligation for instant replies), and emphasize offline tutoring. Many programming assignments need offline meetings for high-quality discussion, or at least a Tencent Meeting.

For course-management questions, first check whether there's a clear policy. If there's ambiguity, discuss with other teachers/TAs to reach consensus.

Don't post announcements or speak casually; think before speaking, and avoid contradicting yourself. Anything we leave in writing must be formal and precise.

Some extreme cases: if a student confronts us emotionally, the TA must avoid confronting back emotionally. Prompt replies more easily lead the TA to also reply in emotion — avoid that. Let your angry words go to the LLM, not to a live person. Don't vent your own life emotions onto students. (E.g. I was just put through the wringer in a group meeting, then answered a student's question with anger, speaking harshly. So: when you're unhappy, don't reply to students. Play a couple of games first, answer when you're in a better mood. Actually not just students — you shouldn't vent emotions onto anyone.)

B. The 50-minute discussion

Pre-arranged questions:

  • What's the future relationship between AI tutors and human TAs?
  • What parts of current TA work can be automated? How?
  • Should the complexity of environment setup be exposed to students?
  • When a student hasn't started on an assignment before the deadline, do you nag? How?
  • In TA work, do you have any moments of great joy, or great frustration?
  • Some TAs feel few people come to offline tutoring/recitations — how do you understand this?
  • About the 5 traps we just mentioned, any thoughts or stories? Or different views?

Questions from students:

  • How to effectively combine the online platform to give feedback on assignment results, and how to gauge tutoring to avoid revealing solutions?
  • If a student's question exceeds the TA's ability (TA doesn't know), how to tell the student?
  • Will working as a teaching assistant help to further continue an academic career at the University?

(TODO: based on the discussion, compile some "answers/views" for these questions here)