Stopping research can be rational

Stopping research is a decision about the remaining commission, not a verdict that the original team was foolish. New evidence, a changed opportunity or an unavailable response can make continued work unattractive even when the initial purchase was reasonable. The review should compare the remaining benefits and avoidable costs using the current action architecture. It should preserve the original charter so that a revision is understandable, while allowing the institution to respond to conditions that were not known when the study began.

The relevant cost at a continuation review is the cost that can still be avoided. A fee already irreversibly paid is not recovered by stopping. A remaining field trial, analytical stage or operating disruption may still be avoidable. A contract can also impose a termination consequence that changes the comparison. The paper does not calculate a particular contract or invent termination terms. It recommends a remaining-cost register that distinguishes sunk commitments from future choices and identifies the evidence needed to value each component.

The current decision can change before the study finishes. A new external result may make the proposed signal redundant, a revised design may move the action boundary, or an approval restriction may remove the response that generated most of the original value. The continuation review should update the prior, menu and payoff perimeter only when the reason is documented. A team should not keep quoting the original EVSI as a justification for every additional research stage after its assumptions have ceased to apply.

The no-policy-change and restricted-action fixtures identify distinct reasons to reconsider a commission. In the former, the current prior is already favourable enough that both branches choose the same action. In the latter, the high-upside response is unavailable. Neither case is a dynamic stopping calculation. They are static comparisons that show how a current decision can have less information value under changed conditions. A real continuation rule would need the timing, remaining evidence process and cost of stopping or continuing to be represented explicitly.

The study may also cease to be feasible. The intended population may be inaccessible, an instrument may not perform in the required conditions, or the remaining window may be too short for credible analysis. Continuing can produce a deliverable that looks complete while no longer answers the commissioning question. A method checkpoint should therefore review the evidence object's relevance and quality as well as progress against a schedule. Completion of planned tasks is not sufficient when the plan has lost its connection to an actionable decision.

A partial finding should not automatically become a valid early signal. The team may have observed a subset of the intended data or an ambiguous result that does not follow the commissioned likelihood model. Acting on that partial result can require a new analysis. The paper's binary engine does not implement sequential statistical monitoring or correct for repeated inspection of accumulating data. A governance decision to review the commission is different from a methodologically valid stopping rule for an experiment. The record should keep these questions separate.

The committee should predeclare material review triggers where possible. These can include a changed action deadline, a removed response, a new evidence source, a failure of study access or a material cost revision. A trigger prompts a review; it does not determine the result automatically. The institution may continue because a broader declared purpose remains worthwhile, redesign because another instrument better fits the current decision, or stop because the remaining value no longer justifies the effort. The reasons should be recorded in the same decision language as the original charter.

Stopping can preserve resources while creating obligations. A partially completed study may require safe closure, disclosure of material findings, data preservation or an operating handback. These conditions belong in an appropriate institutional review. The analyst should not assume stopping is costless merely because future analytical work is cancelled. Nor should it invent a general legal duty from a methodological framework. The relevant owners should identify the obligations that actually apply and include their consequences or constraints in the continuation decision.

The decision should avoid escalation of commitment. An institution can be tempted to finish a study because much has already been spent, because stopping might appear embarrassing or because a sponsor promised a report. These incentives can be acknowledged without making claims about actual people. The practical control is to ask an independent reviewer to assess remaining value and cost against the current decision. Preserving the initial rationale helps learning; it should not bind the team to complete work whose commissioning purpose has materially disappeared.

The opposite error is to stop because a preliminary finding is uncomfortable. A sponsor may prefer to end a study that begins to challenge the desired commitment. A predeclared charter and review path can make that behaviour visible. The review should distinguish lost usefulness from an adverse but relevant result. Evidence that makes the committee choose Hold may be exactly the response that justified the original commission. Disliking that outcome is not a methodological reason to suppress the finding or redefine the study's purpose.

The continuation decision should retain available alternatives. It may be possible to narrow scope to the uncertainty that still controls the action, use already obtained evidence, switch instruments or redirect the study to a later declared decision. Each alternative needs its own cost, feasibility and claim scope. A report cannot simply call an incomplete original study a complete new one. The revised deliverable should say what changed and which findings are still supported. The institution should maintain the distinction between reusing valid evidence and retroactively inventing an objective.

A positive gross information value is not sufficient to continue at any price. The expensive-signal fixture shows the basic acquisition logic: a signal can improve the policy while its cost exceeds that improvement. At a continuation review, the relevant price is the supported remaining avoidable cost under the current perimeter. The app can compare a supplied revised aggregate cost in a new explicit fixture, but it does not determine that cost or a sequential continuation strategy. The record should identify that distinction before relying on the calculation.

The completion disposition should follow what remains usable. A hypothetical commission stopped after loss of data access may leave a valid specification and no accepted signal. A commission narrowed after a changed action window may leave a useful technical finding for a later decision but no value claim for the original response. A commission finished with an adverse accepted signal can fulfil its purpose by supporting the predeclared withdrawal or hold. Preserve each evidence object with its actual status and withdraw only the inference it no longer supports. This prevents stopped from becoming an empirical failure state and prevents uncomfortable from becoming a reason to erase valid evidence.

A continuation review needs paired reasons for continuing and stopping. In the hypothetical no-policy-change branch, ask whether another legitimate purpose remains and what it would cost to fulfil. In the restricted-action branch, ask whether restoring the response is a feasible separate intervention, rather than assume the research will create authority. In the expensive-signal branch, compare the remaining avoidable burden with the decision still available, leaving sunk spending outside the argument for continuation. These are governance dispositions; they supply no adaptive experimental stopping rule. Their practical contribution is to make both a stop and a continuation answerable to the current purpose, instead of favouring whichever result preserves the original commission.