How can AI help West Virginia evaluators write teacher observations under Policy 5310?
How can AI help West Virginia evaluators write teacher observations under Policy 5310?
Most of what I write about state evaluation systems starts with a warning: the framework changed, and the internet hasn't noticed. West Virginia is different. Policy 5310 has been stable, every county uses the same rubrics, and there is no local instrument to reconcile. If you are a principal in Kanawha County and I am a principal in Mingo County, we are rating on the same fourteen elements with the same four words.
So the problem in West Virginia isn't that AI gives you a stale answer. It's that AI gives you a confident answer built on the wrong bar.
Here is the specific thing it gets wrong, and once you see it you can't unsee it.
In West Virginia, Distinguished means with students
Read down the Distinguished column of the WV rubrics across Standards 1, 2, and 3, and the same phrase keeps returning:
1.2 — collaborates with others, including students, to design instruction and assessment
1.3 — collaborates with students to design and use a variety of assessments
2.2 — collaborates with students to establish an effective classroom management system
2.3 — establishes, with students, clear criteria for high-quality work
3.1 — collaborates with students to use an extensive variety of instructional strategies
3.2 — facilitates student-led learning activities
3.3 — collaborates with students and others to make instructional decisions
That is not an accident of drafting. That is the whole architecture.
Accomplished is the teacher doing the work well. Distinguished is the teacher doing the work with students.
It is not the same practice performed harder. It is not more differentiation. It is not "really excellent teaching." A teacher can run the most sophisticated, best-scaffolded, most responsive lesson in the county and still be Accomplished, because they designed all of it themselves. And WVDE's own training materials confirm it, listing "Strong collaboration with students" and "Evidence of student-led learning" among the Distinguished markers.
Now think about what a general-purpose AI tool does with your notes. It has read the entire internet on teacher evaluation, and the internet on teacher evaluation is overwhelmingly Danielson. In Danielson, Distinguished means more sophisticated, more individualized, more responsive. So it reaches for that meaning, sees a strong lesson, and hands you Distinguished. It will do this every single time, in fluent, confident, well-organized prose. And it will be wrong every single time, in a state that requires you to document evidence before you can award that rating.
Five standards, or seven?
Both, and this is the other thing worth getting straight.
There are five West Virginia Professional Teaching Standards:
Curriculum and Planning
The Learner and the Learning Environment
Teaching
Professional Responsibilities for Self-Renewal
Professional Responsibilities for School and Community
The self-reflection you file by October 1 covers those five. But WVDE's standards page, right after listing them, adds a sentence that is easy to read past: to review the seven standards for evaluation, see the Evaluation Rubrics for Teachers.
The evaluation runs on seven. The extra two:
Standard 6: Student Learning — whether the work of the teacher results in measurable progress of student learning of the state-approved curricula.
Standard 7: Professional Conduct — policy and procedure, work schedule, attendance, and respect.
Five for reflection. Seven for evaluation. If you have ever wondered why a vendor's West Virginia page lists five standards, that's why — they found the standards document, published it, and never clicked through.
The sixteen elements
The rubric is written at the element level, and elements are what you rate. This matters more than it sounds like it does, and I'll come back to why.
StandardElements1 — Curriculum and Planning1.1 subject matter knowledge · 1.2 standards-driven instruction · 1.3 balanced assessment2 — Learner and Learning Environment2.1 unique characteristics of learners · 2.2 safe and appropriate environment · 2.3 learner-centered culture3 — Teaching3.1 research-based strategies · 3.2 motivates and engages · 3.3 adjusts instruction4 — Self-Renewal4.1 professional development · 4.2 collaborative learning5 — School and Community5.1 school-wide efforts · 5.2 families and community · 5.3 practices and policies6 — Student Learning6.1 measurable progress7 — Professional Conduct7.1 conduct in law, policy and procedure
Sixteen elements, seven standards.
Accomplished is not a C
West Virginia is unusually direct about where teachers are supposed to land, and it says so in the rubric document itself:
The Professional Teaching Standards recognize performance at the Accomplished level to be meritorious as well as rigorous and of high quality. Accomplished performance is expected to be the most frequently recognized level; performance at the Distinguished will likely occur on occasion.
That's a state pre-empting rating inflation in writing, and it's rarer than it should be.
Two more instructions in the same passage are worth holding onto:
Rate current, actual practice — not a future expectation. The rubric says this twice, and specifically flags that it matters most when you're choosing between two adjoining levels. The temptation, when a teacher is almost there, is to rate where they're heading. Policy 5310 says rate where they are.
Preponderance of the evidence. That's the standard. Not one moment, not the worst thing you saw, not the best.
And from WVDE's guidance elsewhere: evaluators will not use checklists to determine ratings. West Virginia has ruled out the mechanical approach explicitly, which is a thing worth knowing about any tool that offers to hand you a score.
Where Emerging and Accomplished actually separate
This is the part that eats your evening, and it's why the element-level structure matters.
The lines between adjacent levels in West Virginia often turn on a single word. Miss the word, miss the rating. A few that catch people:
Element 2.3 — learner-centered culture. Emerging: the teacher sets criteria for high-quality work. Accomplished: the teacher sets and communicates clear criteria. Same activity, one extra verb, different rating.
Element 2.3 again. Emerging: engages students in learning. Accomplished: engages students in active learning. One adjective.
Element 3.3 — adjusting instruction. Emerging: recognizes missed opportunities to modify instruction. Accomplished: modifies instruction when need is apparent. This one is almost poignant — Emerging is the teacher who knows, in the post-conference, exactly where they should have turned. That's real growth. It's also not Accomplished yet.
Element 3.3 again. Emerging: examines student data. Accomplished: analyzes student data to make instructional decisions. Examining is looking. Analyzing is deciding something.
Element 2.1 — unique characteristics of learners. Emerging: differentiated activities for some students. Accomplished: differentiated activities for students. The whole rating hinges on "some."
Element 1.3 — balanced assessment. Emerging: shares assessment data with students. Accomplished: shares data and provides timely feedback. Data without feedback is a gradebook.
Element 5.2 — families. Emerging: responds appropriately to contact from families. Accomplished: seeks relevant knowledge of the family. The difference between answering the phone and picking it up.
I've been on both sides of these conversations for eighteen years, and I'll say plainly: at 9pm, on your fourth write-up, you are not holding fourteen elements' worth of one-word distinctions in your head. Nobody is. That's not a competence problem. That's a working memory problem, and it's exactly the kind of problem a tool should solve.
Where the evidence actually comes from
Seven standards, three different sources. This is where I want to be careful, because it's where most tools quietly overclaim.
What you watch: Standards 2 and 3, plus the delivery side of Standard 1. Roughly seven elements. Observations run not less than 30 minutes each.
What gets shown to you: Standards 4 and 5, plus the planning side of Standard 1. Written plans, assessment design, PD records, PLC participation, communication logs.
Neither: Standards 6 and 7.
Standard 6 is settled by the Student Learning Goals you filed by November 1 — pre and post assessments, two data points in time, measurable and comparable across classrooms, rigorous and tied to the WVCCRS. K-3 teachers must include a goal addressing reading and math proficiency. The rubric asks whether evidence from multiple measures consistently validates progress. Nothing in a lesson touches that.
Standard 7 is attendance records, schedule adherence, and a pattern of professional interaction. The Unsatisfactory descriptor is explicitly a pattern, not an incident — a single sharp exchange on a Tuesday is not a rating. And Standard 7 runs on its own scale: Distinguished, Meets Standard, Below Standard, Unsatisfactory. Not the four levels you use everywhere else.
One more thing that changes the picture, and I don't think it gets said enough: WVDE defines educator evidence as observable practices and/or artifacts selected by the educator and/or the evaluator. In West Virginia the teacher can bring their own evidence. On Standards 4 and 5 they usually are the one bringing it. Which means those standards were never a unilateral call by the evaluator in the first place.
The requirement most tools don't know exists
WVDE says it directly: evidence must be noted in the system to support a Distinguished rating. That evidence can live in the self-reflection, the summative evaluation, or both — but it has to be there.
Sit with the implication. A Distinguished with nothing documented behind it isn't generous. It's a problem waiting for a grievance, and the problem belongs to the evaluator who wrote it. Now recall that a general-purpose chatbot's default failure mode in West Virginia is handing out Distinguished it doesn't understand the bar for. Those two facts meet badly.
The flip side of that requirement is the best thing in the whole policy. WVDE's guidance on constructive feedback says it:
is necessary to improve educator and student performance; uses language from the rubric and includes specific ways to get to a different result; is specific and actionable.
Read that again. The state is describing, in its own words, exactly the job: put the rubric's language in front of the teacher and tell them concretely how to get to the next level. That's not a marketing claim anyone had to invent. It's a specification.
Where AI genuinely helps — and where it doesn't
It helps with the language. You know what you saw. Rendering it in the rubric's vocabulary, at the element level, with the right one-word distinctions intact, is the drafting work — and drafting is what these tools are actually for.
It helps with the distinctions. Sets versus sets and communicates. Examines versus analyzes. Recognizes missed opportunities versus modifies. A tool anchored to the rubric holds all fourteen elements' worth of those lines at 9pm. You don't have to.
It helps with consistency. Same rubrics, four evaluators, one building. Drift isn't a character flaw, it's arithmetic. Anchoring every draft to the same language narrows the spread — and in a state where every county rates on the same instrument, consistency is the entire game.
It helps with the evidence trail. If Distinguished requires documented evidence, drafting the evidence alongside the rating is not a nicety.
It does not know your teacher. Preponderance of the evidence means evidence across a year. A model sees one set of notes.
It cannot tell you whether that practice is who they are. Authenticity, embeddedness, whether the collaboration with students was real or staged for you — that's judgment from having been in the room, and having been in it before.
It cannot give you the rating. It gives you a draft rating. You accept it, change it, or throw it out. Policy 5310 rules out checklists precisely because this is supposed to be a professional judgment, and no tool gets to make it.
What to write down
The draft can only work with what you captured. A few notes that make the write-up dramatically easier in West Virginia specifically:
Write down who designed it. This is the Distinguished question, and you will not reconstruct it later. Did the teacher set the criteria, or did the class? Did the teacher pick the strategy, or did students have a say? Did students help build the management system? A single note — "kids wrote the rubric with her in September, she referenced it" — is the difference between Accomplished and Distinguished across half the elements.
Write down the verb. Did she set criteria or set and communicate them? Did he examine the data or analyze it into a decision? Did she recognize the missed opportunity in the post-conference, or modify in the moment?
Write down "some" or "students." For 2.1, differentiation for some students is Emerging and differentiation for students is Accomplished. That's a count, and counts don't survive to the evening.
Write down the direction of the family contact. Responding to a parent is Emerging. Seeking the family out is Accomplished.
Write down what you couldn't see. The honest write-up names its own gaps. That's not weakness — under preponderance of the evidence, it's accuracy.
Common scoring mistakes in West Virginia
Reading Distinguished as Danielson's Distinguished. The big one. It's about collaboration with students and teacher leadership, not sophistication.
Treating Accomplished as a middling score. The state calls it meritorious, rigorous, and high quality, and expects it to be the most common rating. Rating a strong teacher Distinguished to avoid "insulting" them with Accomplished misreads the whole scale.
Rating the future. The rubric says current, actual practice, and says it twice. A teacher who is almost there is not there.
Rating standards instead of elements. The rubric is element-level. So is the policy: unsatisfactory performance on a single element cannot initiate a Corrective Action Plan. That distinction only exists if you rated at the element level to begin with.
Confusing the observation with the evaluation. Observations are Requirement Four. The Summative Evaluation is Requirement Five, and it covers all seven standards on evidence gathered across the year. Your write-up is an input, not the output.
Forgetting the teacher gets a say. The teacher's signature denotes the evaluator reviewed the evaluation with them. It does not imply concurrence. They may append a statement within five working days.
The rest of the year, briefly
Orientation — annually, at the start of the employment period or within two weeks of a reporting date.
Self-reflection — on or before October 1, on the first five standards, rating current actual practice.
Student Learning Goals — on or before November 1. Measurable, comparable across classrooms, rigorous, based on identified student needs including performance gaps.
Observations — Initial progression: four, two scheduled and two not, spaced across November 1, January 1, March 1, and May 1. Intermediate: two, one scheduled and one not, by November 1 and May 1. Advanced: not required but recommended, and the teacher can request one.
Summative — finalized at the conclusion of the conference, by the teacher's last day or June 15, whichever comes first.
When it goes wrong — a Focused Support Plan runs nine weeks. If that produces inadequate progress and an evaluation shows unsatisfactory performance on one or more standards, a Corrective Action Plan follows: 18 weeks, with at least one observation completed before it starts. If evidence doesn't demonstrate adequate progress at 18 weeks, termination for unsatisfactory performance follows.
How EvalScribe handles West Virginia
EvalScribe is an AI-powered evaluation tool built by a practicing K-12 teacher.
Standards 1 through 5 are built in — all fourteen elements, with language across the four performance levels, the look-fors an observer watches for, and the distinctions that separate adjacent ratings. Drawn from the rubric itself, not from generic teaching heuristics. EvalScribe drafts to the one-word lines, treats Accomplished as the meritorious professional standard West Virginia says it is, and holds Distinguished to the collaborate-with-students bar the rubric actually describes.
You observe the way you already do. Capture notes by typing, dictating, or photographing your handwriting — Smart Scan OCR handles the handwriting, including pedagogical vocabulary that trips up generic OCR. EvalScribe maps your evidence to the relevant elements, drafts best-fit ratings with the evidence behind them, and writes feedback using language from the rubric with specific ways to get to a different result. You review, edit anything, export a clean PDF.
Honest scope, and I want to be plain about it.
Standards 6 and 7 are not in the tool. The test we apply is simple: can you hand EvalScribe the evidence? For Standards 1 through 5 you can — observation notes, or an artifact you can photograph or type. For Standard 6 there is nothing to hand over; it's pre and post measures against goals filed in November, judged across a year. For Standard 7 it's attendance records and a pattern of conduct over months. Both are real standards. Both are on your summative. We're not going to pretend a lesson tells us anything about either one.
EvalScribe does not assign your ratings — every one is a draft you accept, change, or discard. It does not produce your Summative Evaluation. It does not write Student Learning Goals or score their attainment. And it can't know what happened in the eleven months you didn't have notes for, which is what preponderance of the evidence actually means.
On privacy: observation notes are processed transiently on Microsoft Azure and are not retained on our servers. Evaluation records live locally on your device. Nothing is used to train public AI models. Worth noting in West Virginia specifically — your evaluation record lives in the state platform and the teacher has a right to append a statement to it, so where a draft lives before it gets there is not a small question.
Three evaluations are free. The West Virginia page is at evalscribe.com/westvirginia. If you're leading a rollout across a county rather than a building — evaluator consistency across fourteen elements, access codes, a Data Processing Agreement — that's at evalscribe.com/district.
FAQ
Does West Virginia evaluate on five standards or seven? Seven. There are five West Virginia Professional Teaching Standards, and the self-reflection covers those five. The evaluation adds Standard 6 (Student Learning) and Standard 7 (Professional Conduct).
What does Distinguished mean in West Virginia? Collaborating with students on design, criteria, strategies, and decisions — plus contributing to the professional learning of others through teacher leadership. It is not Danielson's Distinguished, and assuming it is will inflate every rating you write.
Is Accomplished a good rating? Yes. The rubrics call it meritorious, rigorous, and of high quality, and say it's expected to be the most frequently recognized level.
Why doesn't EvalScribe draft Standards 6 and 7? Because you can't hand us the evidence. Standard 6 is year-long goal data. Standard 7 is records and a pattern of conduct — and it runs on a different scale.
Do I have to document evidence for a Distinguished rating? Yes. WVDE requires evidence noted in the system to support Distinguished. It can live in the self-reflection, the summative, or both.
How many observations are required? Initial progression: four. Intermediate: two. Advanced: not required but recommended, and the teacher may request one. Not less than 30 minutes each.
Can a teacher submit their own evidence? Yes. WVDE defines educator evidence as practices or artifacts selected by the educator and/or the evaluator.
Does signing the evaluation mean the teacher agrees with it? No. The signature denotes the evaluator reviewed it with them. The teacher may append a statement as an addendum within five working days.
Can one bad element start a Corrective Action Plan? No. Unsatisfactory performance on a single element cannot initiate a CAP.
Can AI write my evaluation for me? It can draft the write-up. It can't watch the lesson, weigh a year of evidence, or decide whether the collaboration you saw was real. Policy 5310 rules out checklists for exactly that reason — the judgment is the job.
References
Evaluation Rubrics for Teachers, West Virginia Board of Education / WVDE, 2025-2026 — all seven standards, all sixteen elements, and the performance-level descriptors.
WVBE Policy 5310, Performance Evaluation of Professional Personnel and Athletic Coaches (126CSR142) — the governing rule.
West Virginia Professional Teaching Standards, WVDE — the first five standards in depth.
Performance Evaluation of Teachers, WVDE Office of District and School Improvement — timelines, progressions, plans for improvement, and the evidence requirements.
W. Va. Code §18A-2-8 — reasons for suspension or dismissal.
Verify any citation against the current WVDE publication before relying on it for a board or grievance conversation. Policies get revised.
