Step 8 of 11
Principles: what you hold to while it runs
Six rules for making sound engagement choices. They are what separates the same mode done well from the same mode done badly.
A principle here is a standing control policy. Nelson and Narens give the shape: a meta level monitors what the object level is doing and issues control decisions back down. Monitoring tells you what is happening; a principle is what you have decided in advance to do about it.
This is the layer that explains a fact the mode list cannot. Two people can both be doing Production Assistant — handing a draft to the model and taking back a better one — and end up with entirely different artifacts. One keeps authorship of the argument and their own sentence rhythms; the other gets back a homogenised default and ships it. Same mode, same tool, same task. The principles are the difference.
Which is also why a mode count is not a quality measure, and why the framework needs this layer to avoid becoming a scoring rubric for behaviour it cannot interpret.
The six, and what each rests on
Two of these draw on sources already introduced on the theory step, used for a narrower job here.
Supports Authorship
Psychological ownership
What it established. People come to feel that something is theirs through three routes — controlling it, coming to know it intimately, and investing themselves in it — and that feeling changes how responsibly they treat it.
What it licenses here. Why authorship is a principle rather than a legal question about credit. All three routes run through the parts of the work AI most readily absorbs, which predicts that ownership weakens quietly as delegation grows unless it is deliberately held.
Pierce, J. L., Kostova, T., & Dirks, K. T. (2003). The state of psychological ownership: Integrating and extending a century of research. Review of General Psychology, 7(1), 84–107. Source
Supports Productive Friction
Desirable difficulties
What it established. Practice conditions that slow acquisition and feel worse at the time often produce better long-term retention and transfer than conditions that make performance smooth immediately.
What it licenses here. The principle's whole content: not all friction is waste, and removing it is a choice with a cost that shows up later. It does not license treating friction as good in itself — the finding is about specific manipulations, and an obstacle that teaches nothing is just an obstacle.
Schmidt, R. A., & Bjork, R. A. (1992). New conceptualizations of practice: Common principles in three paradigms suggest new concepts for training. Psychological Science, 3(4), 207–218. Source
Supports Calibrated Trust
Trust as appropriate reliance
What it established. The goal is not more trust or less trust in an automated system but trust matched to its actual reliability. Both over-reliance and under-reliance are failures, and each has its own costs.
What it licenses here. Why the model refuses to treat reliance as a deficit or independence as a virtue. Calibration is the target, and it is the reason Verification is a mode rather than a personality trait.
Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human Factors, 46(1), 50–80. Source
Supports Dialogue, not Transaction
Initiative belongs to both parties
What it established. Useful systems are not vending machines. Either party can take the initiative, and the interaction is a negotiated sequence rather than a request and a response.
What it licenses here. Why the single-shot prompt is treated as a missed opportunity rather than merely an inefficient one: the part of the exchange where the person's own question gets reshaped only exists across turns.
Horvitz, E. (1999). Principles of mixed-initiative user interfaces. Proceedings of CHI '99, 159–166. Source
Supports Voice Preservation
Writing as a process with a writer in it
What it established. Writing is not transcription of finished thought. It is a recursive process of planning, translating and reviewing in which goals are discovered and revised while writing, not before it.
What it licenses here. Why voice is a principle and not a stylistic preference. If goals are formed during composition, then handing composition over hands over some of the goal-setting too — the loss is not decorative. It does not license the claim that AI-assisted writing always costs voice; it identifies where the cost would occur.
Flower, L., & Hayes, J. R. (1981). A cognitive process theory of writing. College Composition and Communication, 32(4), 365–387. Source
Supports Reflection
Metacognitive monitoring
What it established. People can monitor and regulate their own cognition, and this capacity develops, can be taught, and is frequently poor when left to itself.
What it licenses here. The principle that compounds the others: examining your own engagement as a site of learning is what turns any of the five above into something durable. Together with Schön it is the mechanism behind the Becoming loop.
Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new area of cognitive–developmental inquiry. American Psychologist, 34(10), 906–911. Source
Dispositions are what you bring, principles are what you hold to, and the eight modes are what actually shows up in the transcript. The next step covers the fifth layer — what all of this is supposed to move — and the one after it reports what the research has found, including where the framework's own predictions did not hold.
Quick check
You are working on a hard problem. Which is true of the six principles?