B.5.MPC:10.3 - Why notation and understanding remain part of the work

An expression can help a practitioner perform a construction. Macbeth’s account of paper-and-pencil reasoning explains how a diagram or inscription can participate in the reasoning, including by allowing the same content to be analysed in more than one way. Dutilh Novaes examines formal languages as cognitive tools whose use depends on learned abilities to read and manipulate signs. These accounts support the operative expression step in :4.3. They do not establish that a notation improves every task or that a human learning effect transfers unchanged to AI. See Macbeth, 2011 and Dutilh Novaes, 2012, §§3.2, 5.2 and 6.1.

For this Method, the practical consequence is precise. Naming the same count N in several expressions is useful only while its participant and operation remain recoverable. A gear graph helps reason about closed contact paths because its edge meaning and traversal rules are available. The four card states distinguish occupancy from reservation throughout entry and exit, where an undifferentiated “not free” count gives only a bound. These are changes to what the expression helps someone do, not merely choices of appearance.

A second alternative is to make one formal language carry the whole inquiry. This can help when a mature language expresses the required physical, mathematical and execution distinctions and its users can work with it. If it cannot express a necessary distinction, use another representation or develop the language. Retaining interpretable correspondences allows several forms to contribute without assuming that one form already covers the whole problem.

Levenchuk’s 2012 robotics account describes difficulty combining familiar speed calculations, several distance quantities, program expressions and physical timing. It motivates changing the question while retaining the interpreted relations in :4.3, and examining the timing of observation and execution in :4.5. The account is a historical report of a particular learning situation. The resulting Method here is a methodological synthesis.

AI can reduce the cost of obtaining a calculation, candidate proof or explanation while leaving the choice and interpretation of the receiving question open. Klowden and Tao discuss the difference between a formally checked statement, its intended meaning and the understanding that enables further use. Section :4.8 turns that distinction into a contribution question: who can recover the decisive connection and adapt it when the premise changes? This is a capability to arrange, not an assertion that every participant must reproduce every proof. See Mathematical Methods and Human Thought in the Age of AI, 2026, §4.