AI Agent Teaching Applications Workshop (3): From Self-Practice to Classroom Formative Assessment

Smart Rainforest Industry Talent Development Program | This page is the handout for Workshop (3), compiled from planning documents; the official announcement for each cohort governs actual dates.

Course Information

This workshop runs two identical sessions — please register for one of them. Actual seat availability and the registration link follow the official announcement.

Agenda:

1. Series Positioning and Context

This is the third and final workshop of the "AI Agent Teaching Applications Workshop" series, designed for university faculty without an engineering background. Across three progressive, hands-on sessions, the series helps teachers see AI moving beyond a simple chat or prompting tool into an AI Agent that can help operate computer and cloud workflows. The core method transfers across tools: state the goal, context, constraints, and completion criteria clearly; check the results; then revise in small iterations.

This session builds on the question set and self-practice tool produced in Workshop (2), bringing after-class self-practice back into the classroom as a formative-assessment prototype where students log in to answer in class and the teacher receives submission records and scores. The collaboration method established in Workshop (1) — state things clearly, then check and revise — extends here to the fuller "log in, answer, submit, grade, aggregate" teaching scenario.

2. Workshop Overview

The self-practice tool built in Workshop (2) is well suited to after-class self-practice, but there's no way to guarantee students aren't leaning on AI to get the answers while answering, so it isn't suitable as a formal assessment on its own. This session brings that same question set back into the classroom: students log in and answer in class, right then and there, submitting for the system to auto-grade on the spot; the teacher side receives each student's login time, submission time, score, and per-question results.

The teacher dashboard deliberately stays minimal: it only shows necessary information at two points — login and submission — with no live progress tracking and no complex monitoring view. The point is to let participants finish a working prototype with an AI Agent within a single workshop session, rather than getting stuck on a more complex feature like real-time monitoring. Once the whole class has finished, the teacher can review overall per-question accuracy and common errors, and hand that summary to AI to help compile a simple learning-analysis report as a reference for the next lesson.

This session also starts from conversation: collaborate with an AI Agent to state, in plain language, what the student side should show, what the teacher side should show, and what data should be recorded at login and at submission, and let the AI decide how to build it. When the result doesn't match expectations, practice asking the AI to fix it using only "what I'm seeing" and "what I want instead" — not editing the code yourself, and not restating the whole requirement from scratch.

The goal of this session is likewise not to build a formal exam platform with identity verification and anti-cheating mechanisms, but to complete a formative-assessment prototype that can actually be tried out in the classroom. Participants will experience firsthand extending the collaboration method from Workshop (1) and the question set and tool from Workshop (2) into the fuller "log in, answer in class, submit, grade, aggregate" teaching scenario — while also being clear about what this deliberately minimal version can and can't do yet.

3. Who Should Attend

• Faculty who attended Workshops (1) and (2), or who already have a multiple-choice question set (questions, options, and answers) ready.
• Faculty who want to bring after-class self-practice into the classroom, so students can log in and answer, then submit on the spot.
• Faculty who want to see student answering and submission status and preliminary scores, catching common misconceptions early.
• Faculty comfortable with "usable first, complete later" — prioritizing a deliberately minimal, classroom-only minimum viable version over building a formal testing platform right away.

4. Learning Outcomes

After completing this session, you will be able to:

1. Collaborate with an AI Agent to clearly state a requirement for "students log in to answer in class, the teacher receives submission records," and ask the AI for brief follow-up fixes when the result doesn't match expectations.
2. Convert Workshop (2)'s multiple-choice question set into a classroom-answering version where students log in, answer in sequence, submit, and get auto-graded.
3. Build a deliberately minimal teacher dashboard showing student login time, submission time, score, and per-question results, and explain why it deliberately skips live tracking and complex monitoring.
4. Once the whole class has finished, aggregate per-question accuracy and common errors, and collaborate with AI to compile a simple learning-analysis report as a reference for the next lesson.
5. Explain what this tool can and can't do, and use an acceptance checklist to confirm the full login → answer → submit → grade → aggregate flow meets the original teaching requirement.

5. Focus Points and Minimum Viable Features

Workshop focus points:

• Compare the differences between after-class self-practice and classroom formative assessment.
• Convert Workshop (2)'s question set into a classroom-answering version.
• Design the basic data flow for login, answering, and post-submission.
• Build a teacher dashboard: submission, timing, score, per-question results.
• Aggregate class-wide accuracy and common errors, and use AI to produce a learning analysis.

Minimum viable features (this session's acceptance baseline):

• Students can log in and answer in sequence during class; the system auto-grades on submission and immediately shows the individual score.
• The teacher side shows each student's login time, submission time, score, and per-question results.
• The teacher dashboard deliberately skips live progress tracking and complex monitoring, showing necessary information only at login and submission.
• Once the whole class has submitted, the teacher can see class-level statistics such as per-question accuracy and common errors.
• The class-level statistics can be handed to AI to help compile a simple learning-analysis report.
• Questions, answers, and the mapping to learning objectives must still be checked by the teacher; question-set data and the web page code stay cleanly separated, making it easy to swap in different materials or edit questions later.

6. Preparation Checklist

• Bring a laptop and confirm it has a working Python interpreter installed (if not, search the Microsoft Store for "Python 3.12" and confirm with python --version).
• Bring the question-set data completed in Workshop (2) (or at least one usable multiple-choice question set) as the source material for this session's exam server; you may also start from the sample question bank the instructor provides.
• Bring at least one extra mobile device (phone or tablet) to simulate a student logging in and answering, to test against your own teaching laptop.
• Questions, answers, and learning objectives must still be checked by the teacher; do not use student personal data, full grades, or other sensitive data.
• If you attended Workshops (1) and (2), bring or organize the AI collaboration task card, question set, and self-practice tool you completed then.

To review Workshops (1) and (2)'s environment and setup basics beforehand:

In the live session, you'll collaborate with an AI Agent to build "student login and classroom answering" and "the teacher dashboard" from scratch, deliberately skipping live tracking and complex monitoring for now.

7. Scope, Positioning, and Limitations

This session's output is positioned as a lightweight classroom testing tool, useful for pop quizzes or formative assessment with an existing question bank, not as a formal online testing platform with anti-cheating or identity-verification mechanisms, and not suited to high-stakes, grade-bearing tests.

• The teacher dashboard is deliberately minimal — no live progress tracking, no complex monitoring — showing necessary information only at login and submission. A fuller real-time monitoring version requires further development after the session.
• Neither the answering page nor the dashboard currently has password protection: anyone who knows the URL can open them, including the teacher's view. Only give students the answering URL; keep the teacher URL to yourself and don't share it publicly.
• Do not expose the server to the public internet (e.g. via port forwarding) — it's for classroom LAN use only.
• Questions, answers, and the mapping to learning objectives must all be checked by the teacher — AI output cannot be adopted automatically; do not paste student personal data or full grades into AI conversations.
• If the question bank contains copyright-sensitive past-exam material (e.g. from the Technological and Vocational Education Joint College Entrance Examination Center), confirm you have the legal right to use and display it before running this tool.
• If formal assessment, long-term cross-class use, personal-data handling, databases, permissions, live monitoring, or cloud deployment are needed later, plan data governance, security, accessibility, operations, and institutional procedures separately.

8. Expected Takeaways

• A multiple-choice question set converted into a classroom-answering version.
• A prototype where students log in to answer in class, submit, and get auto-graded.
• A deliberately minimal teacher dashboard that shows necessary information only at login and submission.
• A class-wide learning-analysis report, aggregating per-question accuracy and common errors with AI's help.
• An AI Agent collaboration method — state the requirement, then check and follow up — you can reuse for other teaching tools.

9. Series Wrap-up and Next Steps

The three-workshop "AI Agent Teaching Applications Workshop" series concludes here: Workshop (1) practiced stating a vague teaching idea clearly and collaborating with an AI Agent on small, checkable tasks; Workshop (2) turned that skill into a question set and a self-practice tool students can use; Workshop (3) (this session) brought it back into the classroom as a formative-assessment prototype where students log in to answer in class and the teacher receives submission records and scores. The core method across all three is really just one thing: state the goal, context, constraints, and completion criteria clearly; check the results; revise; and iterate in small steps — a method that transfers not just to building testing tools, but to any teaching or administrative task you want to hand to an AI Agent later.

There is no further workshop scheduled to follow this session. If you later want to upgrade this session's deliberately minimal, non-live prototype into a full version where the whole class connects at once and the teacher sees progress live, you can continue this session's collaboration method — keep talking with an AI Agent and extend the features step by step. Please keep the code and question-bank data you produce today, and feel free to keep extending this workflow for your own subjects and classes.