V.O.I.C.E.

Methodology

A finding is only as good as the method behind it. This page states the methodological standards every V.O.I.C.E. study must meet — and which of those standards are live today versus still being built.

Every V.O.I.C.E. study record publishes its full methodology as part of the record itself: the research design, the sample frame, the achieved sample size, weighting, the margin of error, the response rate, how responses were collected, and in which languages. These are not footnotes. If a study can’t document its methods, it doesn’t get published.

Open polls and statistically designed research are different things, and they are never presented as the same. A voluntary poll — where anyone who happens to show up can answer — can be interesting, but it cannot tell you what a population thinks. Only a study with a defined sample frame and a documented design can support that kind of claim. The Research Library labels which one you’re looking at, and the two are never combined into a single estimate.

Uncertainty is part of the finding. Every estimate is published with its margin of error or equivalent measure of uncertainty, plus a plain-language statement of what the data cannot say. A number without its uncertainty is a slogan, not a statistic.

As the platform grows, study designs will be pre-registered — published before fieldwork begins — so the plan can’t be quietly rewritten to fit the results. That safeguard, and independent methodological audits, are committed and listed below as proposed until they exist.

Controls in this area

Each safeguard below carries an honest status: proposed means committed but not yet built, implemented means live today, and verified means checked by an independent party.

  • Full methodology published with every study

    Implemented

    Design, sample frame, sample size, weighting, margin of error, response rate, collection mode, and languages appear on every study record — not in a footnote.

    Where to check it: See any record in the Research Library, e.g. its Methodology section.

  • Open polls and designed research are labeled as different things

    Implemented

    Voluntary polls and statistically designed research are never mixed or presented as equivalent. Each carries a label saying which one it is.

    Where to check it: Research Library demonstration records and the homepage integrity note.

  • No representativeness claims without design and evidence

    Implemented

    V.O.I.C.E. will not describe any finding as representative of a population unless the sampling design supports it and the evidence is published.

    Where to check it: Stated as policy on the homepage and enforced in study records.

  • Findings ship with their analysis method

    Implemented

    Published findings say how they were produced: how many responses were analyzed, by what method, and when the analysis was completed.

    Where to check it: The 'About these findings' block on every study record with recorded analysis.

  • No dialogue surface, by architecture

    Implemented

    Issues have no comment, reply, reaction, follow, or message surface. The database schema contains no tables for them — this is not a missing feature but the product: the only action is voting.

    Where to check it: The Issues section: issue pages expose only the vote widget (or endorse widget while proposed). See database/init/005_social.sql.

  • Study designs pre-registered before fieldwork

    Proposed

    Research questions, sampling plan, and analysis plan will be published before data collection begins, so the plan can't be rewritten to fit the results.

  • Independent methodological audit

    Proposed

    A qualified outside reviewer will audit study designs and published methods on a regular schedule.