Instructional Practice with AI

Critical Thinking: Human-in-the-Loop

In this context, critical thinking in action means always treating GenAI outputs as "rough drafts" that require your professional expertise to refine, culturally responsive vetting to protect against historical bias, and instructional alignment to ensure student success. Integrating generative AI into your teaching requires a shift from being a "content consumer" to a "critical editor." This human-in-the-loop approach ensures that technology serves as a co-pilot rather than an autopilot. Human-centered AI use involves highly interactive, iterative work with an AI tool to produce valuable outputs that have been fully vetted by a subject matter expert: you. In practice, this means you must always double-check and verify every AI-generated output for accuracy and pedagogical soundness. The market is currently flooded with generated worksheets and lesson plans, and the vast majority do not meet the rigorous standards CPS holds for High-Quality Curriculum (HQC). When pulling from these external sources, it is your responsibility to carefully review their alignment against CPS HQC standards.

Women Working on Computer

Utilizing AI Meaningfully

Teacher and students In this section, we provide you with some specific examples of how generative AI could be used in your instructional practice. Always keep the human-in-the-loop standard in mind: you should never copy and paste straight from a chatbot into teacher or student-facing documents. Adhering to that, there are several places in your administrative and student-facing processes where generative AI could be a valuable tool.

Administrative Process Use Cases

In your administrative or teacher-facing tasks, generative AI can help you spend less time on busy work so you can spend more time on high-level planning around your students’ needs.

When deciding whether generative AI might be of use to you, ask yourself the following questions:

Is there a task that is repetitive?
Is there a task that can be made easier by quickly creating a very rough draft?
Is there a task where a lengthy document or resource needs to be reviewed against a complex set of criteria?

The following are initial use cases to think about:

Writing emails and letters to students and families is key to maintaining an engaged and transparent classroom environment. Generative AI is particularly effective at taking a body of text and tailoring it to a certain audience or to having a certain tone, as well as translating text. It is also effective at taking a list of bullet points and a desired format (email, Google Classroom announcement, etc.), and synthesizing them into a well-structured message.

Human-in-the-Loop Warnings:

Do not send AI-generated or enhanced communications to families without proof-reading the material.
If using AI to translate communications for multilingual families, be aware that AI often misses cultural idioms or local dialect nuances, which can lead to unintended formalities or misunderstandings.

Examples:

Creating a Gemini Notebook (formerly NotebookLM) that takes your lesson documents from the week, upcoming dates of note or deadlines, and your notes and converts them into a consistently formatted newsletter to share with students and families.
Using Gemini to translate a permission slip into Spanish and Polish, then sharing the draft with a bilingual staff member to ensure the tone remains welcoming and culturally appropriate.
Using Gemini Notebook to create an infographic or visual aid for a message you’re sending out to students and/or families.

Generative AI can help you create a rubric for an assignment with benchmarks of your choice based on the learning goals and/or standards your class is working on. Rubrics can be key to ensuring that students have a clear understanding of your expectations, and that your grading is conducted as objectively as possible.

Human-in-the-Loop Warnings:

Make sure rubric grading criteria is aligned with assessment expectations from lesson TFGs and Unit Scope and Sequence Plans.
Because AI is trained on historical data, it can inadvertently mirror systemic biases or outdated "deficit-based" language that has historically underserved different student populations. Be vigilant in ensuring AI doesn't apply stereotypical descriptors or unfair assumptions based on student demographics.

Examples:

Using Gemini Notebook (NotebookLM) to generate a rubric for a multimedia class project, adapting grading criteria based on the form of media submitted.
Asking an AI to "crosswalk" your draft rubric against Universal Design for Learning guidelines to ensure there are multiple pathways for students to demonstrate proficiency and mastery.

Drafting progress reports, incident logs, or meeting minutes can be a time-intensive burden on an educator's schedule. Generative AI can act as a drafting tool to help structure your observations into professional, concise, and objective reports. By providing the AI with raw data or bulleted notes, you can ensure that the final document follows a logical flow and meets institutional formatting standards.

Human-in-the-Loop Warnings:

Personally Identifying Information (PII) can only be entered in approved products with data share agreements
Because AI is trained on historical data, it can inadvertently mirror systemic biases or outdated "deficit-based" language that has historically underserved different student populations. Be vigilant in ensuring AI doesn't apply stereotypical descriptors or unfair assumptions based on student demographics.

Examples:

Inputting raw observation notes from a classroom walkthrough into Gemini to draft a structured professional development feedback report for a peer.
Using Gemini Notebook (formerly NotebookLM) to help synthesize three weeks of attendance and participation data into a brief, narrative summary for a departmental progress review.

Are you interested in dabbling in a bit of AI-assisted coding? Is there a repetitive task in your Google Suite that takes up too much of your time? Consider using generative AI to help you make automatic scripts in your Google Docs, Drive, Sheets, and Slides.

Human-in-the-Loop Warnings:

Do not test AI-generated code on large datasets or on outside stakeholders.
Ensure you maintain human-in-the-loop standards. There should be a clear audit trail of anything a script does automatically for you, and scripts that directly modify anything facing students or other stakeholders should be thoroughly tested and require manual approval to run.

Examples:

Using Gemini to create an automation script that auto replies to emails when a specific folder reaches a certain capacity.
Using an AI-generated Apps Script to automatically generate a "Certificate of Achievement" PDF whenever a student hits a specific milestone in a Google Sheet tracker.

Student-Facing Materials Use Cases

When using AI as a "thought partner"–especially in a co-teaching or collaborative planning environment–Generative AI can support the implementation of high-quality curriculum with integrity by offering options for adapting and accelerating materials in ways that resonate with their students’ diverse needs and backgrounds, maintaining the balance between consistency and responsiveness. Curricular implementation with integrity means staying true to the critical elements that make a curriculum high quality while encouraging the necessary adaptations in instructional approach, supplemental scaffolding, and enrichment that tailor the curriculum to students’ needs. Integrity & Learning Acceleration: Shared Definitions & Research offers you a deeper dive into implementing curriculum with integrity versus fidelity.

Our Definition of High-Quality Curriculum

Chicago Public Schools built Skyline, the District’s custom PK-12 digital curriculum, in English Language Arts, Artes del Lenguaje Español, Mathematics, Science, Social Science, World Languages (Spanish, French), college and career readiness, and AI and digital literacy to provide teachers with high-quality curriculum that meets the District’s six-part definition:

When using generative AI in this context, ask yourself the following questions:

Am I preserving the three critical elements within each lesson?
1. Backwards-planning from and administering end-of-unit assessments without modification, except per IEPs and/or language proficiency levels
2. Teaching towards standards-aligned daily lesson learning targets (both content and language targets) as written, maintaining the intended level of rigor
3. Using the culturally and linguistically responsive complex texts and tasks central to the lesson at the intended level of rigor
Am I creating temporary scaffolds that maintain rigor? Instead of modifying complex texts and tasks, scaffolds provide entry points for all students to access learning at the intended level of rigor. When planning scaffolds, we first leverage embedded supports inside of the HQ curriculum (linguistic scaffolds, Universal Design for Learning Strategies, mini-lessons, etc.) before introducing external supports like generative AI tools.
Am I generating scaffolds with my students in mind? Consider students’ strengths, needs, identities, cultures, backgrounds, and languages as you adapt their instructional approach and consider supplemental scaffolding or enrichment.

The following are initial use cases to think about:

One of the most exciting use cases for generative AI in instructional practice is the ability to differentiate a lesson plan or assignment for various learners. Differentiation is an intricate process, where teachers have to present different content, learning processes, learning products, and learning environments based on individual student goals, levels, and interests (Tomlinson & Moon, 2013). The primary focus of differentiation–especially in a Tier 1, whole class environment–is increasing access and creating entry points to on-grade level instruction, which all students have a right to. Skyline or other learning objects can be scaffolded to target different reading levels, students struggling with specific standards, students ready for an additional challenge, and more.

Human-in-the-Loop Warnings:

PII can only be entered in approved products with data share agreements.
Maintain rigor. Ensure that "scaling down" a text for readability does not strip away the complex concepts or critical thinking required by the grade-level standard.

Examples:

Use AI to create group roles to help ensure that ML (EL) students at all ACCESS levels can fully engage in a meaningful discussion with their peers
Use NotebookLM to create a graphic organizer (not necessarily a scaled down/lower reading level text) that will help an ML (EL) student better understand the complex text.
Use AI to generate "Hint Cards" for a set of math word problems. The AI can create three versions of the same problem: one with a visual starter, one with a step-by-step checklist, and one "challenge" version that requires applying the concept to a new context.
Use AI to create a supplemental assignment for a student who is struggling with a specific standard within the unit.

Generative AI tools are getting better and better at creating infographics, diagrams, images, stories, and even audio. One exciting use case for gen AI is to augment your own materials with captivating visuals, clarifying diagrams, or audio engagement options.

Human-in-the-Loop Warnings:

Check the content of generated images and diagrams to ensure instructional integrity.
AI image generators often struggle with representation bias, frequently defaulting to stereotypical depictions of gender, race, or ability. Always vet generated visuals to ensure they reflect the true diversity of your classroom.

Examples:

Using Gemini Notebook (NotebookLM), create an interactive podcast version of the textbook chapter students are reading for homework.
Have Google’s nanobanana create images that support the text your class is reading.
Use Gemini Notebook to create an interactive study guide that students can prompt for answers to their questions based on textbook material.

Similar to communications for students and families, generative AI can help you workshop the tone with which you provide feedback to students.

Human-in-the-Loop Warnings:

PII can only be entered in approved products with data share agreements.
If feedback was generated with the assistance of AI, be transparent with your students.
Use AI to structure your thoughts, but ensure the final feedback contains specific references to classroom moments that only you and the student share.

Examples:

Have Gemini create an overall summary based on the comments you’ve left on a student’s essay.
Feeding your "shorthand" grading notes into an AI to expand them into full, encouraging sentences.

AI can be used to generate basic and intricate activities and worksheets based on a lesson plan and a set of learning objectives.

Human-in-the-Loop Warnings:

Always ensure that the output aligns with High Quality Instructional Material (HQIM) standards.
AI may suggest activities that are engaging but "off-map" for your specific curriculum. Always check that the generated activity fits into the Unit Plan.

Examples:

Use AI to brainstorm and draft a fill-in-the-blank vocabulary worksheet.
Prompting an AI to generate five hook ideas for a new activity on Ecosystems that specifically utilize local landmarks or nearby parks to make the lesson place-based.

Avoiding AI Meaningfully

Teacher and student This section details some specific examples of how generative AI should not be used in your instructional practice. Avoiding AI meaningfully entails critically thinking about where AI is and isn’t appropriate in your practice as opposed to adopting or rejecting it in its entirety.

Administrative Processes

The rapid evolution of AI technology has introduced significant data security and legal complexities. Our priority remains enabling productive work within a secure and compliant environment, while recognizing and valuing the exploration and innovation happening District-wide. Unapproved, third-party AI products are blocked on the CPS network to maintain our robust AI Governance, protect our District, mitigate risk, and ensure responsible adoption. Approved tools have entered into an enterprise data share agreement, meaning they have a contract with CPS agreeing not to train their models on or store any data from interactions with their product–ensuring student privacy. Do not use non-approved tools in your instructional practice. Read more in the CPS AI Access Overview.

: GenAI detection software is a category of tools marketed as capable of identifying potential AI plagiarism or AI-generated work, but staff and administrators are strongly encouraged to avoid using these tools. These tools frequently produce false positives–incorrectly flagging student work–as well as false negatives–missing AI-generated work. For example, current AI detection software has been found to disproportionately flag work completed by English Learners (ELs) due to linguistic differences or patterns that the software misinterprets as AI-generated content. Overreliance on these tools could lead to unfounded accusations of academic dishonesty, erode trust between teachers and students, and unfairly penalize certain groups of students. When ensuring that students are maintaining academic integrity, prioritize a holistic assessment of student work–considering multiple factors like the writing process, individual student progress, and direct conversations–rather than relying on GenAI detection software. Sources on these findings and proposed alternatives can be found in the Recommended Readings section.
: The rapid evolution of AI technology has introduced significant data security and legal complexities. Our priority remains enabling productive work within a secure and compliant environment, while recognizing and valuing the exploration and innovation happening District-wide. Unapproved, third-party AI products are blocked on the CPS network to maintain our robust AI Governance, protect our District, mitigate risk, and ensure responsible adoption. Approved tools have entered into an enterprise data share agreement, meaning they have a contract with CPS agreeing not to train their models on or store any data from interactions with their product–ensuring student privacy. Do not use non-approved tools in your instructional practice. Read more in the CPS AI Access Overview.

Student-Facing Materials


: When utilizing Generative AI for creating new content—text, images, audio, code, etc.—your ability to ensure quality and accuracy is limited by your expertise or resources available to verify the generated material. Therefore, a foundational principle for the ethical and practical use of GenAI is to strictly avoid the creation of materials that fall outside the scope of your existing knowledge base or that of your immediate support team. Specifically, you should not use GenAI to produce materials that you cannot competently or effectively evaluate and verify for accuracy, completeness, and adherence to necessary standards.

Overview

Mission, Vision, & Purpose

AI Literacy

AI Principles

All Guidance

AI General Guidance for All Stakeholders

AI Guidance for Educators and Staff

AI Guidance for Students

AI Guidance for Families

AI Guidance for Vendors

CPS AI Library

Instructional GenAI Playbook

Toolkit

Glossary

AI Tools

Recommended Readings

Overview

All Guidance

AI Guidance for Educators and Staff

AI Guidance for Educators and Staff

CPS AI Library

Toolkit

Glossary

Glossary

Critical Thinking: Human-in-the-Loop

Utilizing AI Meaningfully

Administrative Process Use Cases

Student-Facing Materials Use Cases

Our Definition of High-Quality Curriculum

Avoiding AI Meaningfully

Administrative Processes

Student-Facing Materials