Professional learning · Coaches, Teachers, Leaders, Staff · 25 minutes
Never Paste This
A fast sort and a de-identification drill so data teams know exactly what student information must never go into an AI tool, and how to ask the question they need without it.
Data teams are the place where the temptation is strongest: a spreadsheet of scores, a chatbot that can "find the patterns," and fifteen minutes left in the meeting. This short opener, sized for the start of a PLC or data meeting, has teams sort realistic data items into three categories, then rewrite a set of raw data-team notes so the instructional question survives and the students become unrecognizable. Teams leave with a one-page "before you paste" check they can tape to the wall.
In this packet
- Handout A: What Goes In? Sort (card sort)
- Handout B: Raw Data-Team Notes (reading)
- Handout C: Before You Paste Checklist (checklist)
- Facilitator key (last page)
TCEA ELE indicators
- C3.3 I am aware of various data collection and analysis tools.
- AI2.1 I know how to apply ethical principles when using or developing AI for education.
- C3.1 I know how to use data to inform instructional decisions.
Full facilitator guide: mglearn.github.io/eles/activities/never-paste-this.html
Handout A for Never Paste This
What Goes In? Sort
Cut apart. Sort each strip. If your pair disagrees, place it face down in the middle for the debrief. Your district's policy takes priority over this key.
Never into an AI tool
Only de-identified, in a district-approved tool
Fine in an approved tool
Handout A for Never Paste This
What Goes In? Sort: cards to sort
Cut apart the strips. Sort each one onto the mat and be ready to explain why.
A spreadsheet of student names, ID numbers, and reading screener scores.
An IEP goal page, with the student's name blacked out but the disability category and campus visible.
A note that says, "The only 4th grader at our campus who uses a wheelchair is falling behind in math."
A counselor's note about a student's family situation that explains recent absences.
Student essays with names removed, to get suggestions for common writing weaknesses across the class.
A class list with first names only and each student's score on a unit test.
The percentage of the class that scored below 70 on each question of a unit test.
The text of a unit test you wrote, to get ideas for reteaching the questions students missed most.
"Suggest small-group activities for 3rd graders who can decode single-syllable words but struggle with multisyllabic words."
A photo of a whiteboard from a data meeting that lists students by name in intervention tiers.
Anonymous survey comments from students about the school climate, some mentioning specific teachers and incidents.
The state's published description of a reading standard, to generate practice question ideas.
Handout B for Never Paste This
Raw Data-Team Notes
Fictional notes from a 3rd grade data meeting at Oak Ridge Elementary in Cypress Bend ISD. Strike every direct and indirect identifier in red. Then rewrite the team's question at the bottom so it needs no student data.
1Beginning-of-year reading screener review, 3rd grade, Ms. Tran's and Mr. Oyelaran's classes.
2Jaylen Brooks (student ID 4471902, born March 3) is below benchmark on the fall screener. Decodes single-syllable words well, but struggles with multisyllabic words. Receives dyslexia services with Ms. Hale on Tuesdays and Thursdays. His mom mentioned at the conference that the family is going through a divorce.
3Priya S. is below benchmark, but she moved here from Brazil in August and is the only 3rd grader in the newcomer program this year. Her oral retell in Portuguese was strong. Same multisyllabic pattern as Jaylen.
4Marcus and Dani (Mr. Oyelaran's class, table 3) are both just below benchmark with the same multisyllabic pattern. Marcus was absent 9 days last spring for medical reasons.
5Team question as written in the notes: "Paste all of this into the chatbot and ask what intervention each of these four kids needs and whether Jaylen's home situation is affecting his reading."
6Your rewrite of the question, with no student data:
Handout C for Never Paste This
Before You Paste Checklist
Read this aloud before anyone opens an AI tool in a data meeting. Every answer should be Yes.
| Yes | No | |
|---|---|---|
| The tool is on our district's approved list for this kind of use. | ||
| We've asked ourselves whether this question needs any student data at all. | ||
| No names, ID numbers, birthdates, photos, or initials are included. | ||
| No detail could single out a student to someone who knows our campus ("the only one who…"). | ||
| No disability, health, family, discipline, or counseling information is included. | ||
| Any student work has been checked line by line for personal details. | ||
| We're sharing aggregated results, not individual rows, wherever possible. | ||
| A person on the team will verify every suggestion before it affects a student. | ||
| We could explain exactly what we pasted to a family and feel fine about it. |
Notes
Facilitator only for Never Paste This
Answer key and notes
Handout A: What Goes In? Sort
| Item | Sort | Why |
|---|---|---|
| A spreadsheet of student names, ID numbers, and reading screener scores. | Never into an AI tool | Direct identifiers plus academic records. |
| An IEP goal page, with the student's name blacked out but the disability category and campus visible. | Never into an AI tool | Disability information with campus details can identify a child; special education records need the strictest handling. |
| A note that says, "The only 4th grader at our campus who uses a wheelchair is falling behind in math." | Never into an AI tool | No name, but the details identify the student to anyone who knows the campus. |
| A counselor's note about a student's family situation that explains recent absences. | Never into an AI tool | Sensitive personal and family information; it has no place in a tool. |
| Student essays with names removed, to get suggestions for common writing weaknesses across the class. | Only de-identified, in a district-approved tool | Student work can contain personal details; strip them, check each essay, and use only an approved tool. |
| A class list with first names only and each student's score on a unit test. | Only de-identified, in a district-approved tool | First names plus a class are identifying; replace names with numbers, and prefer summarized results. |
| The percentage of the class that scored below 70 on each question of a unit test. | Fine in an approved tool | Aggregated item data with no individual students; still use an approved tool. |
| The text of a unit test you wrote, to get ideas for reteaching the questions students missed most. | Fine in an approved tool | Your own materials and no student data. |
| "Suggest small-group activities for 3rd graders who can decode single-syllable words but struggle with multisyllabic words." | Fine in an approved tool | A purely instructional question; no student data needed. |
| A photo of a whiteboard from a data meeting that lists students by name in intervention tiers. | Never into an AI tool | Images carry names just like text does. |
| Anonymous survey comments from students about the school climate, some mentioning specific teachers and incidents. | Only de-identified, in a district-approved tool | Free-text comments often contain names and events; scrub them first and follow district survey rules. |
| The state's published description of a reading standard, to generate practice question ideas. | Fine in an approved tool | Public information; the output still needs a teacher's review. |