Students need context when their first use CUNY CAIL Sandbox, and the instructor should be clear with them about what this tool is, what it can and cannot do, and how you expect them to use it in your course. This page outlines a first-week onboarding plan you can adapt to any discipline before introducing the tool into your classroom.
- Create the models and knowledge bases students will use (see the Design & Test section)
- Set visibility to Limited and share them with your course group
- Test each model with the kinds of questions your students will ask
- Write a brief AI use policy for your syllabus (see the template below)
Students arrive with wildly different assumptions about AI, while others have never touched it. Some students may believe AI is cheating, while others believe it can write their papers for them. A clear policy sets expectations.
Your policy should address:
- What is allowed: Which assignments permit AI use? Which do not?
- How to cite AI use: What disclosure is expected when students use AI in their work?
- What counts as the student's work: Where is the line between using AI as a tool and submitting AI output as your own?
- Privacy: What should students avoid sharing with the model? (Personal data, other students' work, confidential materials, etc.)
Template:
This course uses the CUNY CAIL Sandbox for specific assignments marked [AI-Assisted]. For these assignments, you may use the course model to brainstorm, get feedback, or explore ideas. You must disclose AI use in your submission and describe how you used it. All final writing, analysis, and argumentation must be your own. Do not paste other students' work or personal information into any AI tool. For assignments not marked [AI-Assisted], AI use is not permitted.
Make sure to adapt the template to specific context as courses will be more permissive or restrictive depending on the instructor’s design.
Frame the Sandbox in your course context. Students need to understand why you chose this tool for this course.
Discussion Prompts:
- "How many of you have used ChatGPT or a similar AI tool? What did you use it for?"
- "What do you think AI tools are good at? What are they bad at?"
- "Why might a [discipline] course use AI tools differently than a [other discipline] course?"
Walk students through the login process. Project your screen or share a link to the Getting Started page.
- Navigate to chat.ailab.gc.cuny.edu
- Click Sign in with CUNY Login
- Authenticate with CUNY credentials
- Locate the course model in the model selector dropdown
Give students a low-stakes prompt to try. This builds comfort with the interface.
Starter Prompt: "Ask the course model to explain one concept from last week's reading that you found confusing. Then ask a follow-up question about its response."
Walk the roomand answer questions if any. Let students discover the interface at their own pace.
Once students are comfortable with the interface, introduce critical engagement.
- Ask students to prompt the model with a factual question from the course
- Copy the model's response
- Identify three specific claims in the response
- Verify each claim using course readings or library databases
- Report back: Which claims checked out? Which did not? Which could you not verify?
Why this exercise: It sets the expectation that AI output requires evaluation. Students who do this exercise in week 1 are less likely to treat the model as an authority throughout the semester.
Tip: Save 5 minutes at the end for a whole-class debrief. Ask: "What surprised you?" The answers usually surface the key lesson on their own.
"Is this cheating?" Not when used as directed. The course’s AI use policy defines the boundaries, so using the course model for an [AI-Assisted] assignment accordingly is the same as using a calculator in a math class or a spell-checker in a writing class. However, using it on a non-AI assignmentor submitting AI output as your own work is a direct violation of course policies.
"Can the AI see my other conversations?" Each conversation is separate. The model does not remember previous chats unless you are in the same thread.
"Is CUNY reading my conversations?" The Sandbox is hosted on CUNY infrastructure so the Sandbox administrators can access system logs. Your conversations are not shared with other students or used to train AI models. Other students cannot see your conversations as well unless you are in a group channel. For more information, see the Sandbox privacy documentation.
"What if the model gives wrong information?" It will. AI models generate plausible text, and sometimes that text is inaccurate. That is why this course teaches you to evaluate AI output critically. Treat the model's responses as a starting point for your own thinking, not a finished product.
- Reference the onboarding experience in later assignments.
- Invite the class to revisit and revise the AI use policy mid-semester as students' understanding of AI evolves with use.
- Share interesting student discoveries with the class. When someone finds a creative use or catches an interesting error, make it a teaching moment.
- Check in with students about their AI use every few weeks; ask what is working and what is frustrating
- Adjust model configurations based on how students actually use them
- Update your knowledge bases as the course progresses (add new readings, remove outdated materials)
- Teach@CUNY AI Toolkit — CUNY-specific guidance on teaching with AI
- Getting Started — the student-facing login guide
- Teaching Tips — strategies for integrating AI throughout your course