All activities Professional learning
Instead of a demo and a vote, teacher teams test an emerging AI feature inside a GRITS lesson-study loop, watch students rather than the tool, and decide from student evidence.
Teachers ask "Can we use this?" about a new AI feature almost weekly. The usual answer is a demo and a show of hands. This lab replaces the vote with a structured tryout built on the GRITS lesson study model: the team sets a learning goal from student data, picks the research-based strategy the tool must serve, tries the fictional MarginMate annotation tool's new AI Hints feature as learners, plans a classroom tryout where colleagues observe students instead of the teacher, and then analyzes returned student work from two teams to reach a decision with a rubric. The possible decisions are adopt, adjust and retry, or drop, and adjust is the most common honest answer.
Hook
Project the fictional email from Handout A: a teacher asking whether the department can turn on MarginMate's new AI Hints feature for the argument unit. Ask: "Hands up: yes, no, or not sure." Record the split. Then say: "We just voted on a tool none of us has seen with students. Today we find out what the vote should have been based on."
Facilitator noteEarly votes are usually split by enthusiasm, not evidence. Keep the tally visible and return to it at the end.
Explore
Teams read the tool profile on Handout A and fill in the G and R rows of Handout C. G, Goal setting: what learning gap does the student data show? (The packet gives one: students highlight but do not identify an author's claim or evidence.) R, Research: which proven strategy should the tool serve? Offer choices: explicit modeling and think-alouds, worked examples before independent practice, sentence-level writing, or claim-and-evidence argumentation. Only then do teams try the feature as learners: annotate the printed stand-in text in Handout A without hints, then read the printed AI Hints and note what each hint did to their thinking.
Facilitator noteAsk one question at every table: "Did the hint help you think, or did it do the thinking?" The printed hints include one that misstates the author's purpose. See who catches it.
Model
Model the I phase of GRITS. Explain that in the classroom tryout one teacher teaches while colleagues observe students: what they annotate before and after a hint, whether they reword the hint or copy it, whether anyone questions a hint, and who cannot reach the hint at all. Think aloud as an observer reading a student's screen: "She highlighted the claim first, then opened the hint, then changed her label from 'evidence' to 'claim.' That's a change I can see. I'll note it with a time."
Facilitator noteName the stance: this is not an observation of the teacher. It is a look at whether the tool changes student reasoning.
Create
Teams complete the I row of Handout C: which class, which lesson, the comparison (for example, hints available only after a student's first own annotation, or one section with hints and one without), what observers will record, the privacy check (district contract, data use, student accounts), the access check (screen readers, English learners, students without home devices), and the date the team brings evidence back. Remind teams that the tryout lasts one or two lessons, not a semester.
Facilitator noteIf a team cannot answer the privacy question from the profile, that is itself a finding. The tryout waits until someone confirms the district's data terms.
Apply
Say: "Two teams ran this tryout. Here is what they brought back." Teams read the returned evidence from Team Cypress (Grade 8 ELA) and Team Pecan (Grade 11 U.S. history) on Handout A, using these stems: "I notice students…" · "This work shows a misconception about…" · "The hint helped when… and got in the way when…" Then they score the tool on Handout B. A coach facilitates the decision protocol: (1) clarifying questions only, 3 minutes; (2) evidence round, each person names one piece of student work and what it shows, 6 minutes; (3) concerns round, including privacy and access, 4 minutes; (4) the team places a sticky on Adopt, Adjust and retry, or Drop and states the condition that would change its decision.
Facilitator noteMost teams land on Adjust and retry: hints only after a first attempt, the district data opt-out confirmed, and the screen-reader problem fixed before a second tryout. That is a strong outcome, not a failure to decide.
Debrief
Complete the S row of Handout C: what changes in the lesson before it is retaught, how the team will document the tryout for the PLC, and what it will tell the teacher who sent the email. Then return to the Hook vote. Ask: "What would you vote now, and what evidence moved you?" Close with the coaching question: "What changed in teacher practice or student evidence?"
Facilitator notePoint out that the team's reply to the email now includes conditions and evidence. That reply is a reusable model for the next "Can we use this?" message.
Transfer
Each coach names the next tool or feature teachers have asked about, and drafts the G and R rows of a fresh Handout C for it. Teachers name the lesson where they would host an observed tryout.
Facilitator noteKeep the list of pending tool requests visible in your coaching space. Running one tryout a month is sustainable; evaluating every request is not.
The packet includes a printed stand-in: a short text, the AI Hints the feature would show, and the returned student work from two teams. Participants annotate by hand, read the hints on paper, and make the decision from printed evidence. In a school with no devices for PD, this is the entire lab, and the classroom tryout happens later where the tool is used.
If your district already approves an annotation tool with an AI hint feature, teams can try it on a public-domain text during the Explore step, with the facilitator's account projected rather than personal accounts. After a real classroom tryout, teams post de-identified student work in a shared folder and score the rubric in a shared document so the decision and its evidence stay together for the PLC.
Does it need a screen? Trying the feature live shows how the hint appears, how fast it arrives, and whether it can be ignored, which paper cannot fully show. The decision itself should rest on student work from a classroom tryout, which is why the printed evidence carries the most weight.
What you should be able to see or collect if it worked.
Teacher teams test an emerging AI feature in a lesson study cycle and decide from student evidence using a rubric.
Pick the next tool request on your list and run the G and R conversation with the requesting teacher this week. Schedule one observed tryout lesson within a month, and bring the student work and the team's decision to your PLC.
