RESEARCH PROPOSALRESEARCH PLAN

What repeated across three experiments?

We review the results and counterexamples from weeks one to three alongside the status of earlier Lab experiments to define what the evidence supports.

Document version
Synthesis plan 01
Updated
Design status
Schedule fixed / awaiting prerequisite results
Schedule
2026.10.18–10.26

What we want to learn

Research question and hypothesis

Across the research actually completed during the month, which results repeated and which claims remain unavailable?

Hypothesis

Format effects are more likely to remain conditional on question and exception type than to appear identically for every product.

This is an expected direction, not an established finding.

Why ask this question?

Reading experiments separately can obscure completion counts and shared limits. The synthesis defines decision scope without forcing unlike methods into one average.

What we will compare

Keep the campaign
change the explanation

Organizes the three experiments, counterexamples, and the completed or blocked state of prior Lab research by condition.

Held constantRaw-evidence links · state separation · conditional interpretation · cost check
A

New experiments A and B

Review direct comparisons of content format and exception placement.

B

Replication C and prior studies

Review the new-product replication alongside results or incomplete reasons for EXP-000–003.

Measure each condition separately
Research stateWhat was actually completed?

Separate completed, blocked, undecidable, and unmeasured work with evidence paths.

Practical decisionHow far can it be applied?

Summarize repeated directions, counterexamples, automated-review limits, and replication needs.

Human response and AI output are measured separately. One does not stand in for the other.

Controls

Do not combine results that differ in target, unit, period, service condition, or method; record conditional differences instead.

Example: travel luggage

Place results from A and B beside replication C, while showing studies stopped at a gate, such as EXP-003, separately from effect results.

How we will proceed

Proposed sequence

  1. Reconcile results and raw evidence

    Recheck that completion counts, errors, calculations, costs, and raw paths for A–C match.

  2. Connect prior-study states

    Classify EXP-000–003 as completed, blocked, or unmeasured, recording the last check and resume condition.

  3. Write the public synthesis

    Explain repeated results and application limits plainly and link every research plan and result report.

Draft by October 24, then verify recalculation, sources, and actual costs on October 25–26 before publication.

What we will record

Measures and limits

MeasureMethodLimit
Completed-study countCount only work with execution, raw storage, review, and result writing finishedDo not count scheduling or technical checks as completed research
Repeated directionCompare direction by product and question with replication CDo not average results from different methods
Blockers and resumptionRecord reason, first time, deadline, and resume conditionDo not turn a blocker into a negative result
Actual costAggregate A–C and prior experiment costs by scopeSeparate direct measurement cost from total operations

How we will judge results

Rules fixed in advance

A or B repeats in new products

Keep it as a conditional execution-rule candidate and plan replication in another context

Results are mixed or fail to replicate

Do not create one rule; publish counterexamples and condition differences

Research is blocked or unmeasured

Exclude it from effect conclusions and retain only the resume condition and deadline

Do not promote observations or technical checks because few completed results exist. Do not set a product default before the evidence level and external replication support it.

Decisions before launch

Items not yet fixed

October 24
Synthesis draft and completed/blocked list
October 26
Recalculation, source, and cost review complete
Public scope
Completed results and verifiable limitations
Incomplete work
Link resume conditions and next-check deadlines

Scope of this plan

  • A–C are supplied-material, no-search experiments and do not establish web-citation effects.
  • Automated scoring and AI rechecking remain distinct from human review.
  • One month of internal replication cannot establish a general prescription across markets and services.

Background and sources

Reviewed guidance
and research rationale

Verisca public research notes

Connects weekly plans and result reports to synthesize the month’s actual completion state.

Checked 2026-09-26

Read the full backgroundFull explanation and hypothetical example

Evidence reviewed together

We connect the comparison results from Experiments A and B, the replication with new products in Experiment C, and either the results or the unresolved status of EXP-000 through EXP-003 to their raw evidence.

What we will not combine

We will not average results that differ in question, product, service condition, or method. An answer experiment that does not depend on indexing cannot support claims about web discovery or citation.

Publication rule

Only completed results enter the synthesis after recalculation, source checks, and actual-cost verification. Every unresolved study receives a deadline and a condition for resuming.