AI literacy professional development matrices

Planning frame

The five elements map cleanly onto teacher learning goals: understand AI systems, apply the right tools, evaluate outputs, reason ethically, and translate AI literacy into each subject's standards of evidence and meaning.

1. Architecture and evaluation

What kind of AI system is this? How was it trained? What makes it reliable or unreliable for a local classroom task?

2. Application and tool choice

How do teachers and students pick the right tool for brainstorming, tutoring, feedback, accessibility, research, or production?

3. Output evaluation

How do learners trace claims, verify sources, test reasoning, compare evidence, and decide what to trust?

4. Ethical use

How do schools handle plagiarism, cognitive offload, privacy, equity, environmental impact, deepfakes, bias, and community norms?

5. Disciplinary specificity

What does “good AI use” mean in ELA, math, science, social studies, arts, CTE, world languages, and other domains?

Design principle

Keep AI literacy and assessment integrity related but distinct: one teaches capable judgment with AI; the other redesigns evidence of student thinking.

Kindergarten–Grade 5 matrix

Elementary PD should treat AI literacy as guided language, curiosity, evidence habits, and protected human thinking—not as independent chatbot access for young children.

Grade focus	Systems architecture	Appropriate application	Evaluating outputs	Ethical use	Disciplinary specificity
K–1 Language, play, adult mediation	“AI guesses from patterns” using picture-sort or autocomplete analogies. Difference between people, search engines, calculators, and chatbots. Teacher-facing: what student data should never go into tools.	Teacher uses AI to create differentiated read-aloud questions, picture-word banks, or parent-friendly explanations. Whole-group “ask, listen, decide” demonstrations with teacher as driver. Use AI for accessibility supports only after checking age, privacy, and district approval.	“Does that sound right?” routines: compare AI statements to class books, photos, observations, and teacher knowledge. Spot the silly answer: AI can be fluent and wrong. Use classroom anchor charts for check-with-a-grownup / check-with-a-source.	Protect imagination, drawing, oral storytelling, and productive boredom from instant generated answers. Talk about kindness, privacy, and not pretending computer words are your own. Use environmental impact as “tools use real electricity and water; use them for a reason.”	ELA: AI as a class-made “noticing partner,” not a replacement for voice or story. Math: AI/calculator comparisons after students model with objects. Science: ask AI a question, then test with observation. Social studies: connect claims to class texts, maps, and community knowledge.
2–3 Early inquiry and explanation	Model training as learning from many examples; discuss why missing examples create blind spots. Introduce “input, model, output, human check” as a simple system diagram. Evaluate tools by purpose: helper, storyteller, translator, image generator, quiz maker.	Prompt frames for teacher-guided vocabulary, examples/nonexamples, and revision suggestions. Use AI to generate choices that students must sort, critique, or improve. Practice when not to use AI: first draft thinking, mental math strategies, personal reflection.	Get it in / track it down: turn AI claims into “things to check.” Use two-source checks with library books, databases, and teacher-provided links. Teach students to ask, “What would prove this?”	Distinguish help, copying, and pretending. Discuss bias through whose examples are included or left out. Class norms for using AI with adult permission and naming the help received.	ELA: compare AI-generated and student-authored descriptions for voice and detail. Math: have AI explain a solved problem, then students find the missing step. Science: use AI explanations as hypotheses to test. Social studies: “source, place, time, perspective” mini-routines.
4–5 Upper elementary responsibility	Mini-lessons on training data, pattern prediction, hallucination, and confidence. Compare open chatbot, search result, school-approved tutor, and closed learning app. Teacher PD: local task tests before adopting an AI feature.	Use AI for brainstorming questions, study games, translation support, reading-level alternatives, or feedback checklists. Design “think first, ask second, revise third” routines. Have students document what they asked, what they accepted, and what they rejected.	Claim-evidence-reasoning checks for AI answers. AI citation detective: find whether a quote, fact, or source actually exists. Compare model output with a human-curated source set.	Introduce mature vs immature AI use: using help to learn vs using help to hide. Discuss cognitive offload: which thinking should stay with the learner? Equity and access: not every student has the same tools, accounts, or adult guidance.	ELA: annotate an AI chat transcript like a reading passage. Math: decide whether AI helped reasoning or just answer-getting. Science: test AI claims against data tables and observations. Social studies: corroborate AI summaries with primary/secondary sources.

Middle school matrix: Grades 6–8

Middle school PD can move from adult-mediated exposure toward structured student practice: prompted inquiry, visible process, disciplinary verification, and norms for independent-but-accountable use.

Grade focus	Systems architecture	Appropriate application	Evaluating outputs	Ethical use	Disciplinary specificity
6 Structured exposure	LLM basics: tokens, prediction, training data, retrieval, and why fluent text can be wrong. Compare model, search engine, calculator, and recommendation system. Evaluate a tool with a simple “job interview” checklist tied to class tasks.	Use AI as vocabulary coach, exemplar generator, practice partner, or feedback mirror. Teacher-designed guardrails: approved prompts, source sets, and reflection logs. Tool-choice lesson: when simpler tools or human collaboration are better.	SIFT-for-AI routines: stop, investigate source, find better coverage, trace claims. Students label output as claim, evidence, reasoning, opinion, or unsupported assertion. Practice follow-up prompts that demand evidence rather than praise.	Classroom boundary chart: allowed, ask first, and not allowed. Plagiarism as misrepresentation, not merely tool use. Intro to deepfakes, privacy, bias, and the environmental footprint of unnecessary generation.	ELA: AI as peer-response partner; students judge whether feedback fits author intent. Math: critique AI solution paths and error patterns. Science: compare AI explanations to lab data. Social studies: source AI claims about events and perspectives.
7 Visible thinking	How systems are optimized: why “helpful” can become overconfident or sycophantic. Closed vs open systems: chatbot, embedded app assistant, AI grader, recommender. Bias and benchmark limits: why local evaluation matters.	Use AI to generate counterexamples, alternative explanations, study questions, or debate roles. Students submit prompts, notes, revisions, and rationale—not just final products. Practice “with, without, about AI” rotations in one unit.	AI output audit: source trace, quote verification, reasoning check, and missing perspective check. Students build a correction memo for a flawed AI answer. Compare different tools on the same disciplinary task.	Mature use rubric: purpose, transparency, effort, evidence, reflection. Cognitive offload discussion: what did the tool do, and what did you still learn? Equity/access planning for students with and without premium tools.	ELA: chat transcripts as texts—annotate questions, revisions, and moments of judgment. Math: use AI for strategy comparison only after independent attempt. Science: AI-generated models must predict or explain observed data. Social studies: civic/media literacy around synthetic media and manipulation.
8 Transition to independence	System diagrams for generative AI workflows: prompt, model, optional retrieval/tool, output, human verification. What training data, fine-tuning, and human feedback imply for reliability and bias. Local pilot design: how a team would test an AI feature before trusting it.	Student-created AI use plans for projects: purpose, constraints, prompts, and disclosure. Use AI to rehearse oral defense, create study pathways, or stress-test arguments. Introduce tool-specific affordances: image, voice, coding, tutoring, translation.	Triangulate AI answers with databases, primary sources, calculations, and expert criteria. Use negative results: when the tool fails, students explain what that teaches about the task. Assess the transcript as evidence of thinking.	Policy case studies: cheating, deepfakes, privacy, bias, environmental claims, and social dependence. Students write use disclosures and reflect on what they intentionally did not outsource. Discuss wellbeing: AI companionship, overreliance, and adult visibility.	ELA: authorship, voice, citation, genre, and revision ethics. Math: conceptual understanding vs procedural answer-getting. Science: uncertainty, modeling, and claims-data-reasoning. Social studies: sourcing, corroboration, civic judgment, and perspective.

High school matrix

High school PD should prepare teachers to design authentic tasks, teach sophisticated verification, separate AI literacy from assessment integrity, and define subject-specific standards for ethical use.

Grade focus	Systems architecture	Appropriate application	Evaluating outputs	Ethical use	Disciplinary specificity
9–10 Foundational fluency	LLM architecture at a conceptual level: tokens, embeddings, training, fine-tuning, retrieval, tools, and agents. Evaluate model fit: task, stakes, data privacy, cost, accessibility, latency, and known failure modes. Discriminative vs generative systems, especially for grading, recommendations, and detection.	AI as tutor, critic, interviewer, editor, simulator, translator, coder, or brainstorming partner—with purpose statements. Prompting as a literacy practice: context, constraints, examples, audience, and iteration. Students select tools for the job and defend why AI is or is not appropriate.	Research workflow: get context, track claims, verify sources, synthesize in the student's own frame. Quote and citation verification with external sources. Compare AI outputs across models and explain differences in evidence quality.	Transparency norms: disclosure, process logs, and boundaries for graded work. Plagiarism vs legitimate assistance; offloading vs outsourcing. Environmental impact and vendor transparency: real footprint, contested viral claims, and proportional use.	ELA: conversational authoring, voice, interpretation, citation, and human-authored literature. Math: proof, reasoning chains, counterexamples, and conceptual explanation. Science: model limits, uncertainty, data analysis, and empirical grounding. Social studies: historiography, civic media literacy, source context, and bias.
11–12 Advanced judgment	Model evaluation labs: design prompts, benchmark against local rubrics, record failure cases. Bias, drift, prompt sensitivity, hallucination, sycophancy, and opaque vendor claims. Procurement literacy: what questions to ask before trusting an AI system in school.	Capstone AI-use plans: roles, constraints, audit trail, source policy, and reflection. AI for code review, data exploration, debate prep, design critique, and specialized study modes. Agentic workflows only when students design the work and can explain the chain of actions.	Evidence portfolios: transcript, sources, annotations, corrections, and oral defense. Students identify when AI amplifies bad problem framing. Advanced source triangulation across primary sources, datasets, scholarly work, and domain experts.	Case seminars: AI grading, surveillance, synthetic media, student wellbeing, access tiers, labor, water/energy, and social trust. Community-level ethics vs individual mature-use rubrics. When to refuse AI help to preserve learning, authorship, relationships, or civic responsibility.	ELA/humanities: authorship, originality, rhetoric, interpretation, and ethical otherness. STEM: simulation, coding, proof, lab design, uncertainty, and reproducibility. CTE/arts: client constraints, design process, tool provenance, critique, and portfolio authenticity. World languages: translation, cultural nuance, oral fluency, and communicative purpose.
Teacher teams Course redesign and assessment integrity	Department-level tool audits: what systems are embedded in existing platforms? Common evaluation protocol: task sample, risk analysis, equity impact, validation, bias check, and negative results. Design living guidance rather than one-time tool lists.	Unit redesign sprints: where AI supports access, feedback, practice, or authentic production; where it should be excluded. Model effective AI use in front of students: uncertainty, revision, source-checking, and judgment. Shared prompt libraries tied to learning goals, not generic tricks.	Assess visible thinking: checkpoints, conferences, process notes, demonstrations, and oral defenses. Grade the chat/process when appropriate, not only the polished artifact. Create departmental examples of acceptable, questionable, and unacceptable AI-assisted work.	Separate AI literacy from cheating prevention so PD does not collapse into policing. Policy alignment: privacy, accessibility, equity, student wellbeing, environmental claims, and community values. Response plans for deepfakes, false accusations, and inconsistent access.	Departments define evidence: what counts as rigor in this subject? Build subject-specific rubrics for AI use, disclosure, and verification. Map “thinking with, without, and about AI” into each course sequence.

Crosswalk: possible PD strands

These strands can become separate workshops or a yearlong sequence.

AI systems and local evaluation

From “AI guesses patterns” in elementary to high-school model audits and district procurement questions.

Critical doing with AI

Prompting, revising, comparing, documenting, and reflecting as literacy practices rather than magic commands.

Verifiable inquiry

Get context from AI, then track claims down to sources, data, observations, and disciplinary evidence.

Productive friction

Design when students should think first, use AI second, and preserve the struggle that builds understanding.

Ethics and mature use

Separate community decisions about whether AI is allowed from classroom coaching on mature, transparent use.

Disciplinary AI literacy

Help every department define what AI changes about evidence, explanation, authorship, rigor, and authentic work.

Wiki sources scanned

This page synthesizes themes from these notes in Clay's personal knowledge wiki rather than inventing a generic AI-literacy checklist.

AI literacy has to be taught inside real subjects — disciplinary evidence, rigor, authorship, and subject-specific examples.
Students need to check AI answers against real evidence — AI as provisional context to trace, test, source, and synthesize.
AI literacy requires different kinds of AI interaction — open chatbots, approved assistants, and closed systems carry different responsibilities.
AI tools should be judged by the work they will actually do — local task testing over vendor claims or generic benchmarks.
Education should teach thinking with, without, and about AI — a three-part duty for AI-ready schools.
Schools should start with learning values before choosing AI tools — values-first implementation and teacher-led redesign.
Learning still needs some struggle, even when AI can make things easier — productive friction, visible process, and cognitive boundaries.
AI literacy and assessment integrity need separate workstreams — teach capable AI use while separately redesigning evidence of learning.
AI grading systems need transparency, validation, and bias checks — especially relevant for evaluating systems that assess students.
AI-era media literacy needs emotional resilience too — verification plus resilience against cynicism, manipulation, and deepfakes.
Rushed school AI plans can worsen wellbeing and equity risks — developmental sequencing, vulnerable students, and safeguards.
AI water and energy research page — environmental impact as real but contextual, with proportional use and vendor transparency.