Plot Twist: Beat the Bot by Designing Backward

Long before generative AI entered the conversation, Stephen Covey urged readers to “begin with the end in mind”, to get clear on the destination first, since that clarity is what shows you where you actually stand and keeps every step after that pointed the right way.

Stephen R. Covey, The Seven Habits of Highly Effective People (cited in Wiggins & McTighe, 2005)

Picture a familiar moment: you are revising a writing assignment, and a colleague mentions using ChatGPT with their students. Should you add it, too? Cut it? Let students choose? Before you touch a single AI tool, there is a better first question to ask; one that has nothing to do with AI at all: What do I actually want students to learn, and how will I know they have learned it?

That question is the heart of Backward Design, a course-planning framework developed by Grant Wiggins and Jay McTighe. Instead of starting with content or activities, “what will I have students do?”, Backward Design starts at the destination. You name the outcome first, decide what evidence would prove students reached it, and only then design the activities, materials, and tools that build toward producing that evidence. Applied to generative AI, this reversal is what keeps a promising tool from becoming a distraction: outcome first, evidence second, tool last.

“Outcome first. Evidence second. Tool, AI or otherwise, last.”

Three questions before you touch AI

Chapter 6 of Ludwig and Zakrajsek's The Science of Learning Meets AI distills this into three questions worth pinning above your desk:

1. What am I actually trying to help students accomplish?  
2. What would count as real proof they got there?  
3. Where, if anywhere, does AI genuinely help with that?

Here is the twist that makes this framework genuinely useful rather than just another checklist: if your honest answer to question three is “it won't,” that is not a failure. Deliberately leaving a tool out can be just as thoughtful a design choice as building one in. Some of the most effective teaching moments work precisely because no tool is involved (a spontaneous debate, a hand-raised question, an unassisted first draft). This cuts both ways: Backward design does not push AI into every assignment; it clarifies where AI earns its place and where it does not.

A Progression, not a prescription

Because faculty rarely start from the same place Ludwig and Zakrajsek organize AI adoption into three stages rather than a single formula:

• Adapt: you are refining an assignment you already trust. AI's role here is auditing; flagging where your outcome, evidence, and activity have quietly drifted apart, not inventing something new.
• Create: you are building something new, or closing a known gap. AI becomes a thought partner, generating outcome language or authentic task formats for you to choose among. The pedagogical decision, however, stays yours.
• Embed: backward design has become routine. You check alignment continuously, weekly, or at midterm, using accumulating student evidence instead of relying on instinct alone.

Most instructors finalizing an assignment for the first time will find themselves in Adapt or Create, and that is exactly where the framework expects you to be.

Three traps worth checking for

Even with clear outcomes in hand, it is easy to fall into predictable patterns. Before you finalize an AI-integrated assignment, check it against three guardrails: tool first, goal later (AI is in the assignment because “students are using it anyway,” not because it serves the outcome); assessment mismatch (students use AI for one kind of thinking, like brainstorming, but are graded on something else, like recall); and innovation without intention (AI shows up because it is new and exciting, not because the assignment actually calls for it). In each case, the fix is the same move: return to the outcome, and keep AI only where it demonstrably serves it.

The bottom line

None of this rejects technology outright. It simply insists that purpose comes before the tool, every single time. The goal was never to use AI more; it is to use it on purpose, so that every AI touchpoint in your course can be traced back to the outcome it serves. Whether you are revising a syllabus you have taught for a decade or building a brand new assignment from scratch, the same three questions apply. Start with where you are going, and let that decide what role, if any, AI plays in getting there.

This is not just how Ludwig and Zakrajsek frame it, either. In a 2026 conversation with podcast host Catlin Tucker, Jay McTighe (Wiggins's original co-author) revisited backward design specifically for the AI era, pointing to how instructors themselves can lean on AI to support their own curriculum and assessment design work. McTighe's standard mirrors backward design's own logic: outcome first, evidence second, tool last. The goal has to be understood and earned before AI enters the design process, not outsourced to it from the start.

Try It Yourself: New Self-Paced Module

An asynchronous Instructor Pathway, “Finalizing Your AI-Assignment,” walks you through this exact process. No prior reading required. Bring a draft assignment (or sketch one as you go) and work through roughly 4 hours of exploring the principles of backward design, developing a toolkit for integrating U of A-subscribed generative AI tools into course assignments, and applying feedback from fellow participants to refine and strengthen your assignment design and support student learning.

Resources:

UCATT Instructor Pathways
AI in Teaching
U of A-Supported GenAI Tools

References:

Covey, S. R. (1989). The 7 habits of highly effective people: Powerful lessons in personal change. Free Press.

Ludwig, L. D., & Zakrajsek, T. D. (2026). The Science of Learning Meets AI: A Practical Faculty Guide to Purposeful Integration, Student Engagement, and Ethical Practice. 

Wiggins, G., & McTighe, J. (2005). Understanding by design (Expanded 2nd ed.). Association for Supervision and Curriculum Development.

McTighe, J. (2026, April 21). Backward design in the age of AI: From coverage to deep learning [Interview by C. Tucker]. The Balance. https://catlintucker.com/2026/04/backward-design-in-the-age-of-ai-from-coverage-to-deep-learning-with-jay-mctighe/