Defend conclusions and preserve the limits of the evidence

Try

Your investigation has a result, a report and supporting calculations. A defense asks whether you can explain why the result follows, which assumptions it uses and what happens when one assumption changes. It is not a performance of certainty or a repetition of a prepared abstract.

Begin by making an evidence map: every final claim should point to a proof, a certified computation, an observation or a clearly labeled assumption. Claims with no supporting path should be removed, weakened or left provisional.

Predict

A reviewer changes the loss-channel model by allowing the rates to vary with time. Does the constant-rate recovery formula still identify two constants? Predict which step of the derivation fails before trying to defend the original conclusion unchanged.

Experiment

With time-dependent rates, total stock is M=M₀ exp(−∫(a+b)dt). Total observations reveal cumulative combined loss, not generally two separate rate functions. Channel A loss is ∫a(t)M(t)dt, so a single final channel fraction is a weighted integral rather than one constant ratio.

Defense question	Required response
What remains true?	Total balance and combined cumulative loss relation
What fails?	Recovery of constant a and b from one fraction without constancy
What could repair it?	A justified restricted time-varying family or richer time-resolved observations

This response is stronger than claiming the model is “robust” without calculation. It locates the assumption, derives the changed relation and states what evidence would be needed next.

Understand

A modeling evidence package includes the question contract, sources, derivation, observation model, data provenance, fitting history, validation, uncertainty and sensitivity, reproducible computation, report and revision log. A proof-centered package includes precise claims, definitions, dependencies, complete arguments, counterexamples, exact certificates where used and unresolved questions. A numerical package additionally documents error norms, reference solutions, work budgets and refinement results.

The presentation should select a small number of central claims and explain their mechanisms. Keep the complete derivation and artifacts available for questions. A useful defense challenge changes an input, removes a hypothesis, asks for a counterexample or requests an independent check. The response should add reasoning, not merely repeat the conclusion.

Review status must remain accurate. A submitted artifact records work. Automated exact checks establish the particular claims they test. A human mathematical review assesses specified evidence according to stated criteria. None alone proves universal competence, originality or validity outside the investigated scope. The teaching unit does not itself deliver an independently reviewed capstone; that requires an actual reviewer and recorded feedback.

After review, preserve the defended version and create a revision addressing substantive objections. If an objection remains unresolved, identify the conclusion that must remain provisional. An honest negative or limited result can be a successful mathematical investigation when its reasoning is sound and its scope is clear.

Apply

Deliver a report, a concise presentation and reproducible supporting artifacts. Ask an independent reviewer for at least one unfamiliar mathematical challenge. Record the question, your response, any follow-up calculation and the revision outcome. Include school, college or workplace interpretation only where the evidence supports it; do not convert a synthetic demonstration into an operational recommendation.

Continue

Use the review to choose the next investigation: a stronger theorem, a better observation design, a more accurate solver or a carefully bounded application. Record what was demonstrated and what remains open. Continued inquiry follows from the evidence, not from a completion badge or a claimed page-count threshold.