See it show up: the homework chatbot

AI shows up in different forms. A feed ranks. A thermostat follows rules and sensors. A chatbot generates language.

For the homework chatbot, a careful description is:

• Inputs: your prompt, chat history, maybe school-safe filters
• Task: produce helpful-looking text that matches the request pattern
• Outputs: sentences, lists, sometimes fake-looking citations
• Hidden risk: sounding right while being wrong, outdated, or incomplete

The system is strong at pattern-shaped writing. That is not the same as checking a textbook page, remembering your teacher’s exact explanation, or knowing what your mark scheme wants tonight.

Trap 1: Fluency = truth. Smooth paragraphs are easy to invent. Truth needs a check.

Trap 2: One good answer = always reliable. Ask a second question with a known fact from class. Does it stay solid?

Trap 3: Chatbot = all AI. Your poster should leave room for recommenders, scanners, detectors, and tools that never chat.

Trap 4: If it helps me, the label does not matter. Helpfulness and honesty are different jobs. You can use a tool and still describe it carefully.

Take one chatbot answer and mark each sentence:

• Checkable — I can verify in notes or a trusted source
• Guessy — maybe useful, needs proof
• Style only — filler that sounds academic but adds little

You do not need to reject the whole reply. You need to see which parts earned trust.

Then rewrite your prompt to tighten the job: “List three causes from our worksheet vocabulary only. If unsure, say unsure.” Bounded asks make bounded judgment easier.

On the hallway poster, a fair line might be: “Some AI tools write answers that look human. Looking human is not the same as being checked.”

That sentence teaches classmates the form without hype. It also connects to the course case: your school only wants tools labeled AI when the description stays honest under a teacher’s question, “What does it actually do?”