You designed the AI-assisted writing task carefully: brainstorming and outlining handled by AI so students can focus on the argument. Here's the problem: the messy beginning you just removed was the learning. Not the warm-up, the learning. This episode introduces the Difficulty Audit, a three-category design tool that tells you exactly where AI belongs in a task, and where it cancels the encoding you cannot afford to lose.

This episode draws in part on research in:

• Desirable difficulties and durable encoding (Bjork & Bjork, 2011)
• Productive failure and pre-invention learning effects (Kapur, 2016)
• Cognitive load theory and instructional design (Sweller, 1988)
• Retrieval practice and generation effect (Roediger & Karpicke, 2006)
• Prior knowledge as prerequisite for productive struggle (Kapur & Bielaczyc, 2012)