Capability built: Design a complete unit or module using Claude , from learning outcomes through assessment , with human review gates at every stage and a verification checklist that confirms the teacher, not the tool, made the decisions that matter.

The capstone deliverable has seven components. Each was introduced in an earlier chapter; this is the first time they appear together, connected by a single teaching purpose. These seven components are not independent. They must cohere: the assessment must measure the outcomes; the feedback plan must serve the assessment; the materials must build toward the lessons; the AI-use policy must be consistent with the assessment design. Claude can help draft each component. The coherence check is your job.

The backward-design principle (Wiggins and McTighe, 2005): Begin with what you want students to be able to do, then design activities and assessments that produce and reveal that ability. Do not start from activities. I am designing a three-week argument writing unit for a first-year general education writing course. Students have varied prior writing experience. At the end of this unit, I want students to be able to write a clearly argued, evidence-supported essay on a topic they have researched.

I have five learning outcomes for a three-week argument writing unit. I will paste them below. Please design a lesson-by-lesson sequence (nine class meetings, 75 minutes each) using backward design logic: activities should build toward the outcomes, and each meeting should do assessable cognitive work, not just deliver content. Include what students do before, during, and after each class. Flag any meeting where the cognitive work is lower than I probably want. Claude output (condensed , three representative meetings):

One material needed: The source evaluation checklist mentioned in Meeting 4. Claude prompt , checklist generation:

CAST UDL guidelines (CAST, 2018) frame differentiation as offering multiple means of representation, action and expression, and engagement , not as creating separate tracks for different students. Claude prompt , differentiation review:

A feedback plan answers four questions: When does feedback happen? What form does it take? Who gives it? What must the student do with it (Brookhart, 2017)? For this argument writing unit, please draft a feedback plan that answers: When does feedback happen in the sequence? What form does it take (written, verbal, peer, instructor)? Who gives it? What is the student's required response to feedback? The plan should ensure that feedback happens early enough to improve the final essay , not just on the final draft.

The assessment must measure whether students achieved the learning outcomes. It must not be an artifact test , a test of whether they produced a document , but a learning test (Ambrose et al., 2010). I have five learning outcomes for my argument writing unit. Please design an assessment that measures each outcome directly. The final essay cannot be the only evidence , at least some outcomes must be measured through process evidence, not just product.

The policy must be teachable , specific enough that students know what is and is not allowed, honest about the reasoning, and consistent with the assessment design (Chapter 10). For this argument writing unit, please draft a student AI-use policy that: (1) specifies which parts of the unit students may use AI tools to assist with; (2) specifies which parts they may not; (3) explains the reasoning in plain language; (4) states the disclosure requirement; (5) is consistent with the assessment design.

Before the unit goes live, you run it through a final check. This is the record of the unit's integrity , where Claude helped and where human judgment decided. Outcomes Outcomes specify what students will do, not what will be covered Each outcome is assessable by at least one piece of evidence in the assessment plan Outcomes were written or reviewed by the instructor, not accepted verbatim from Claude

Grant Wiggins and Jay McTighe published Understanding by Design in 1998 (Wiggins and McTighe, 2005). Their argument was simple and durable: start with what you want students to be able to do, then design backward to activities and assessments. A quarter-century later, that logic is exactly what prevents AI-assisted unit design from becoming artifact assembly. The backward design discipline is not about Claude; it is about whether the unit is actually designed for learning.

I have five learning outcomes for a three-week argument writing unit. I will paste them below. Please design a lesson-by-lesson sequence (nine class meetings, 75 minutes each) using backward design logic: activities should build toward the outcomes, and each meeting should do assessable cognitive work, not just deliver content. Include what students do before, during, and after each class. Flag any meeting where the cognitive work is lower than I probably want. Claude output (condensed , three representative meetings):