Constructed Insights
A model for guiding a person from open question to understanding without giving them the answer.
Developed
Mechanism Established
Human Behaviour
When a learner is given a ready answer, even when it is correct and logically follows, it fails to challenge them critically. They do not explore further and cannot recall the same information, or recognize it in a different context. Since they did not work to reach the conclusion, they cannot reliably apply, extend, or defend the conclusion.
Mechanism
The system withholds the conclusion and surfaces the conditions that make it reachable. A tutor working on why empires fall asks what happens when an army has nothing left to conquer. Each answer narrows what remains possible and moves the person from open territory towards a defensible and justificable position. The person progresses with their own reasoning. When they arrive, the conclusion is discovered rather than delivered. The structure of the sequence is identical to its use in Cognitive Scaffolding. What differs is that the system here holds the destination in advance, knows when the person has reached it, and can recognise a defensible but wrong conclusion as a failure. That knowledge changes how the sequence is operated, not how it appears.
What makes it distinct
Constructed Insight is a model of conclusion ownership where the person arrives at a known answer through their own reasoning. The nearest adjacent model is Cognitive Scaffolding, which uses the same surface behaviour — structured questioning that withholds a conclusion — but toward a destination that does not exist in advance. Because the destination is known here, the system can verify arrival, correct missteps, and close the conversation against a fixed criterion. Cognitive Scaffolding can do none of these and is structurally committed to that limit. A system that already holds the answer and one that does not are architecturally different, even when a transcript of the two would look interchangeable.
Where it breaks
If the person reaches a defensible but wrong conclusion through the sequence, the system has no reliable mechanism to correct without disrupting the structure. The model requires the system to know either the correct conclusion in advance or the criteria by which a good conclusion can be recognised, making it unsuitable for domains where neither condition holds.