### Reflection After completing Parts 2 and 3, I would make several changes to my original problem network. One node that now seems incomplete is “Personalize scaffolding, examples, and representations.” Before this assignment, I treated personalization as an inherently useful way for AI to improve learning. However, personalization is only valuable if the information being adapted is accurate. During my AI audit, most responses were supported by authoritative sources, but the question about whether spring force in a disk-shaped cocking lever always acts at a constant radius was answered incorrectly. This showed me that a confident, personalized explanation can still reinforce a misconception. I would therefore add a node such as “Verify technical claims and communicate uncertainty” as a means of improving the usefulness of AI-generated instruction. This reflects the assignment's emphasis on checking AI against authoritative sources rather than relying on its output alone. I would also revise “Improve design decision quality.” My original network implies that better or faster feedback will fairly directly produce better decisions. After learning the ball screw and toggle-latch mechanisms, I think an important intermediate step is missing: “Develop the ability to evaluate and verify feedback.” AI was particularly helpful for quickly explaining unfamiliar terminology, relationships, and design considerations. For example, it helped me identify ball-screw lead, efficiency, end-fixity types, and toggle-latch holding capacity as topics worth investigating. However, I still had to compare those claims against technical sources before trusting them. The incorrect cocking-lever response in my audit made this limitation especially clear. My chosen problem, “Provide timely, actionable formative feedback,” still seems like the right level of abstraction. I originally selected it because it identifies an educational need without prescribing a particular AI implementation. Parts 2 and 3 reinforced that choice. When learning unfamiliar mechanisms, rapid feedback helped me identify gaps in my understanding and revise my explanations while I was still working. However, I would now refine the problem statement to emphasize timely, actionable, and verifiable formative feedback. The experience showed me that speed alone is not enough; useful engineering feedback must also help students recognize uncertainty, check claims, and make informed revisions. %% #HW01 %%