Skip to the content.

The Next-Move Memo

A short, repeatable record of research judgment, for project work where the answer does not exist yet

Students get practice executing well-defined tasks. They get much less practice deciding where to direct effort when several next questions compete, and that decision is most of what research judgment is. The Next-Move Memo makes the decision visible so an instructor can assess it while the work is still uncertain. It comes from the companion paper Learning to Create with AI, where it sits between passes through the book’s Learning Spiral as a navigation step rather than a seventh discipline.

The memo is useful precisely because the final answer may not yet exist. It also separates AI assistance from ownership: AI may suggest candidate moves or challenge the diagnosis, but the student must identify the bottleneck and defend the choice.

The six rows

Row What the student writes
Current question What are we trying to establish, understand, design, or explain now?
Current bottleneck What presently prevents progress?
Candidate moves What are the two or three strongest things we could do next?
Chosen move Which one will we do now, and why is it more informative than the alternatives?
Result What happened, including failed attempts or negative evidence?
Revised question What should the next pass through the Learning Spiral now address?

Keep it light enough to repeat. A page is too long; six lines to six short paragraphs is the range.

How to assign it

Cadence. One memo at each decision point in a project: after the first exploration, whenever the work stalls, and before any change of direction. Weekly is a good default for a semester project. The Result and Revised question rows are filled in at the next memo, so each memo closes the one before it.

AI use. This is usually an AI as Collaborator (AIC) assignment. The student may use AI to generate or challenge candidate moves and to argue against the stated diagnosis. The bottleneck and the choice are the student’s, and the memo carries a one-line disclosure of how AI was used (students/disclosure-templates.md gives the wording). A student who wants AI’s help running the check can use this prompt, from the paper:

Help me run a Next-Move Check. Do not choose for me until I have stated my own diagnosis. Ask me for the current question, the main bottleneck, and any anomaly or result I cannot explain. Then help me formulate at most three next moves and compare what each would teach me if it succeeds or fails. End by asking me which move I choose and why.

What is not the point. A leverage test is useful when the path is unclear, but creativity also depends on curiosity, play, and occasional inefficient exploration. If students treat every branch as a short-term return on effort, they abandon difficult or anomalous ideas before their value shows. The memo should restore direction, not remove serendipity; a chosen move justified by “I do not understand this result and I want to” is a good memo.

How to grade it

Grade the reasoning behind the selection, not the number or polish of the candidate moves. The questions that discriminate:

Over a semester, improvement in these answers is evidence that the student is becoming less dependent on assigned pathways. That trajectory, across memos, is a better thing to grade than any single memo.

For mentoring conversations

Four questions are often enough: What is the most important uncertainty now? What are the strongest things you could do next? Which one did you choose? Why is it more informative than the alternatives?

Template

Copy this into the assignment.

NEXT-MOVE MEMO            Project: ______________   Date: ________   Memo no.: ___

Current question:

Current bottleneck:

Candidate moves (two or three):
  1.
  2.
  3.

Chosen move, and why it is more informative than the alternatives:

Result (filled in at the next memo, including failed attempts and negative evidence):

Revised question for the next pass:

AI use (one line: which tool, what it did, what you checked):

The assignment templates in assignment-templates.md share the same logic: grade the judgment, not the product.