When formal guidance is missing, AI policy does not disappear. It becomes a thousand classroom decisions.
A teacher decides whether a student may use AI to understand assignment directions. Another decides whether AI-generated feedback supports revision or replaces it. A department debates whether brainstorming counts as assistance or authorship. A principal responds to families who want a clear rule. A district leader tries to write guidance that applies across grade levels, disciplines, learner needs, and tools that keep changing.
None of these decisions is only about permission. Each requires a judgment about learning: What work is the student supposed to do? What difficulty serves that learning? What support provides access? What evidence would show understanding? What must a student be able to explain, defend, revise, or take responsibility for?
This is the work I mean by educator sensemaking. Sensemaking is the practical work of interpreting an ambiguous change and deciding what it means in a specific context. In schools, it happens when educators translate a broad development such as generative AI into judgments about actual students, actual assignments, and actual institutional responsibilities.
The problem is not that educators are making these judgments. Professional judgment is essential to teaching. Rather, the problem is that many educators are being asked to make consequential judgments without sufficiently clear institutional language or support.
An Evidence Arc, Not a Trend Line
The updated K-12 Teacher AI Evidence Arc brings three national teacher-survey snapshots into one view. The first snapshot concerns adoption. RAND reported that roughly one-quarter of surveyed teachers used AI for instructional planning or teaching during the 2023-2024 school year. Adoption varied by subject and school context, and teachers and principals in higher-poverty schools were less likely to report AI use. Principals in higher-poverty schools were also less likely to report that their schools or districts provided AI guidance. RAND’s report establishes that uneven use and uneven support were present early in the public adoption period.
The second snapshot concerns perceived consequences. In my weighted analysis of a Fall 2025 RAND American Teacher Panel public-use file, 68.9% of teachers reported using AI tools during the 2025-2026 school year. In the same survey, 72.5% said AI makes teachers’ jobs easier, while 61.9% said AI makes students’ learning harder.
Those responses do not prove that AI improves teacher work or harms student learning. They are teacher judgments, not measured outcomes. But the asymmetry matters. The same professional community can experience AI as a practical benefit and a pedagogical concern at the same time.
The third snapshot concerns institutional response. In the Winter 2026 public-use file, 55.3% of teachers reported receiving some guidance, formal or informal, about using AI to prepare to teach. Only 9.3% reported formal written guidance for that task. Formal guidance was similarly rare for making assessments, grading or feedback, student-facing supplementation, analyzing learning data, tutoring, and coaching.
Gallup’s published report describes the broader pattern: across ten work tasks, 18% of teachers received formal guidance in at least one area, 48% received only informal guidance, and 34% received no guidance. These surveys are not repeated measures from a single longitudinal study. Their questions, samples, field periods, and denominators differ. The evidence arc should not be read as a clean numerical progression from one year to the next.
Its value is conceptual. It moves from adoption, to perceived consequences, to institutional guidance. Together, the snapshots describe an implementation problem: educators are using AI, noticing tensions between professional efficiency and student learning, and often working with guidance that is informal, incomplete, or absent. That is the setting in which sensemaking occurs.
Educators Are Not Deciding Whether AI Is Good or Bad
School discussions often flatten professional judgment into a position on AI. Is the teacher an adopter or a skeptic? Innovative or resistant? Permissive or strict? Those categories are too crude.
Educators are making more specific decisions:
- Is this support helping a student reach the learning goal, or completing the learning work for the student?
- Does this task require an initial attempt before assistance becomes useful?
- Is the difficulty productive, or is it an exclusionary barrier?
- What choices must remain with the student?
- What does authorship require when some of the language was generated?
- What evidence would allow the student and teacher to know that learning occurred?
The same educator may encourage AI in one context and restrict it in another without being inconsistent. A multilingual learner using translation support to understand science directions is not the same case as a student using AI to generate the scientific explanation being assessed. A student using AI to produce counterarguments and then testing them against evidence is not doing the same intellectual work as a student submitting an AI-generated argument they cannot defend. A teacher using AI to draft routine parent communication is making a different judgment from a teacher using it to evaluate student writing. The category “AI use” is too broad to carry the pedagogical decision. Sensemaking begins when educators move from the tool category to the learning situation.
Four Judgments Beneath the AI Decision
Week 3 argued that academic integrity and learning design ask different questions. Academic integrity helps establish whether students followed the stated conditions for producing work. Learning design asks what students still had to think through. Week 4 turns to the person making that distinction.
Educator sensemaking about AI often involves four connected judgments.
1. A judgment about effort
What effort is educationally productive here?
Some struggle helps students retrieve, interpret, connect, revise, and build durable understanding. Other struggle consumes attention without serving the learning goal. Educators must decide whether AI is reducing noetic friction that students need or removing a barrier that prevented them from engaging in the first place. This decision cannot be made from the tool alone. It depends on the learner, task, timing, and intended capability.
2. A judgment about access
What support allows the student to participate without changing what the task is meant to assess?
AI can support translation, organization, alternate explanations, communication, and accessibility. Those uses may remove exclusionary friction. Yet a support can also cross into substitution when it performs the interpretation, synthesis, or explanation the task was designed to develop. The important question is not whether assistance was provided. Rather, it is whether the assistance preserved access to the learning or replaced the learning.
3. A judgment about authorship
What must the student be able to stand behind?
Authorship is not reducible to who typed each word. It involves responsibility for claims, evidence, decisions, and consequences. A student may use assistance and still demonstrate authorship if they can explain the claim, justify the evidence, identify what changed, respond to criticism, and revise with intention. Fluent language is not enough. The educator has to judge whether the student has entered into a relationship of responsibility with the work.
4. A judgment about evidence
What would make learning visible?
The final artifact may remain important, but it can no longer carry the whole evidentiary burden. Educators may need a brief oral defense, an initial attempt, annotated sources, a revision explanation, a conference, or a transfer task. This is not a call to document every keystroke. It is a call to align evidence with the capability the task claims to develop. These judgments about effort, access, authorship, and evidence are not side issues. They are the practical substance of AI-era teaching.
Sensemaking Looks Different Across Roles
My current dissertation proposal examines K-12 educator and institutional sensemaking across three role-based perspectives. Classroom-facing educators encounter AI through assignment design, student work, feedback, accessibility, and assessment. Their judgments are immediate and situated: what should happen with this learner, in this task, at this point in the sequence?
Building-level administrators encounter AI through instructional coherence, supervision, professional learning, family communication, and schoolwide expectations. Their question is not only what one teacher should do, but how a school can support defensible variation without becoming incoherent.
District and system-level leaders encounter AI through policy, infrastructure, procurement, privacy, governance, curriculum, and professional support. Their decisions shape the range of choices available to classrooms, even when district guidance says little about learning design.
These are different vantage points on the same educational problem. That matters because a rule that appears clear at the system level may become ambiguous in a classroom. A teacher’s effective local adaptation may be difficult to sustain without building-level support. A school may develop strong practices that remain fragile because procurement, pacing, assessment, or policy structures pull in another direction. Sensemaking is distributed across the system, whether the system recognizes it or not.
Professional Judgment Is Not a Substitute for Institutional Responsibility
There is a tempting response to this complexity: trust teachers. Schools should trust educators. But trust without support can become abandonment. When institutions provide only broad principles such as “use AI responsibly,” the difficult decisions do not go away. They are transferred to individual educators. Each teacher must determine what responsible use means for brainstorming, drafting, feedback, assessment, accessibility, tutoring, grading, and student data.
That creates at least three problems. First, it produces inconsistency students and families experience as arbitrary. The same use may be encouraged in one classroom, prohibited in another, and ignored in a third. Second, it creates an equity problem. Educators with more time, stronger professional networks, better access to professional learning, or more supportive leadership can develop more sophisticated practices. Others are left with improvised rules or vendor-provided guidance. Third, it hides institutional choices. A district may claim neutrality while its platforms, pacing expectations, assessment systems, procurement decisions, and professional-learning priorities quietly shape what educators can do. Professional judgment remains necessary. The institutional responsibility is to create conditions in which that judgment can become informed, discussable, and coherent.
Build Shared Judgment, Not Just Shared Rules
Schools need policy, but policy alone will not resolve the contested cases where learning goals, accessibility, authorship, and evidence intersect. A stronger approach builds shared judgment. That work can begin with four moves.
Give educators a common language
Terms such as productive friction, exclusionary friction, cognitive bypass, learning sequence, evidence of thinking, and authorship help educators describe what is otherwise experienced as a vague discomfort. Shared language does not force identical decisions. It makes the reasoning behind different decisions visible.
Examine scenarios, not only principles
Broad statements produce easy agreement. Concrete scenarios reveal the real differences. Should students use AI to brainstorm claims before drafting? To translate directions? To generate feedback? To outline from their own notes? To summarize a source before reading it? To produce counterarguments before a debate? Scenario discussion and card sorting make professional judgment visible. Educators can compare placements, explain what they are protecting, and identify where access support and cognitive bypass become difficult to distinguish.
Tie guidance to tasks and learning goals
“AI is allowed” and “AI is prohibited” are rarely sufficient. Guidance should identify what capability a task develops, what human work must remain visible, where AI may enter the sequence, what supports are legitimate, and what evidence students should provide. This is more demanding than a universal rule. It is also more educationally defensible.
Preserve uncertainty where it is real
Not every question has a stable answer yet. Schools should be able to mark a use as contested, pilot a practice, gather evidence, and revise guidance. False certainty encourages compliance without learning. Structured uncertainty makes professional inquiry possible.
The Question Beneath the Policy
Educator sensemaking is not a temporary stage schools will outgrow once they write an AI policy. Generative systems will keep changing. Their educational meaning will continue to depend on the learner, discipline, purpose, sequence, and institutional context. Professional judgment will remain necessary because teaching is not the application of a static rule to interchangeable situations.
The leadership task is not to eliminate judgment. It is to stop leaving consequential judgment invisible and unsupported. When guidance lags behind use, educators carry the ambiguity. They decide what effort to preserve, what barriers to remove, what authorship requires, and what evidence counts. Those decisions already shape students’ learning experiences.
Schools should make that work discussable, supportable, and coherent.
Week 5 will turn from educator sensemaking to the institutional conditions around it: policy, assessment, professional learning, leadership, infrastructure, and governance. Those conditions do not merely support implementation after a decision is made. They shape which decisions educators can make in the first place.
Explore the evidence: https://minerclass.github.io/k12-ai-evidence-arc-2024-2026/
Read the full Summer Research Series: https://micahminer.com/summer-research-series/
