Schools are being sold AI as a chat box next to the lesson. A chat box is hard to supervise and hard to grade. The useful unit is an activity: a bounded environment where students investigate something, measure it, and explain what they found, with the teacher visible and in charge. We have built two and run them in sessions.
The activities
Fairness Lab. Students give a generative model a neutral prompt and watch what it assumes. Then they change one word and watch the assumption move. The lesson is not "AI is biased". It is "here is where the default came from, and here is the research that predicted it".
Sound Lab. Data becomes sound. A class hears a rising trend or a repeating cycle before anyone draws an axis, then builds the graph to check what they heard.
On-premises, with the boundaries stated
Some schools want AI activities kept closer to their own environment. Delivered by Masterful Creations STEAM Academy using the owner's self-hosted zOvermind platform when private AI infrastructure is part of the workshop. We can also walk through what a school would need to operate something similar.
On-premises is an operating model. It changes where processing happens. It is not an automatic COPPA, FERPA, safety, or curriculum determination, and we will not tell you otherwise.
Read AI in the classroom without student data leaving the building for the longer version.
What participants leave with
- Students can explain one concrete thing about how a generative model behaves, because they measured it themselves.
- Teachers leave with two activities they can run again and a one-page note on data and oversight.
- Administrators get a plain-language read on the on-premises option, including what it does not decide for you.
How a session runs
- The Fairness Lab. Students give a model a neutral prompt, watch it fill in the blanks with defaults, change one word, and see the default move. Research is linked at each step.
- The Sound Lab. A sonification lesson that turns data into sound, so a class hears a pattern before it graphs one.
- Debrief. What the model did, what a person did, and who is accountable for the result.
- For teacher cohorts, a session on oversight: what the instructor controls, what students see, and what is retained.
Questions we get
Does student data leave the building?
For the activities we run, nothing students type goes to a third-party AI service. The sessions run on hardware we operate. That is a technical statement, not a compliance determination. FERPA, COPPA, safety, and curriculum fit stay with the school and its own policy.
Do we need to buy anything?
No. A session needs student devices with a browser. If a school later wants to run AI activities on its own hardware, that is a separate conversation with its own requirements.
What does the teacher do during the session?
Supervises, sees what each student ran and wrote, and leads the debrief. The activity is bounded so there is something specific to supervise.