Sampling protocol for a century-long study of authorial judgment in the scientific objectivity and values literature (Phase 5.1, pre-registered before execution)

Résumé fourni par la source

This is the sampling protocol for a study of whether the visibility of authorial judgment has changed across a century of the literature that argues about scientific objectivity and values. It is deposited before any data are collected. At the moment of deposit no query has been executed, no work has been selected and no text has been coded. The primary frame is journal-based and enumerable. It covers research articles, review articles and essays published between 1926 and 2025 in fourteen sources: Philosophy of Science, the British Journal for the Philosophy of Science, Synthese, Erkenntnis, Studies in History and Philosophy of Science, Social Studies of Science, Isis, Philosophy of the Social Sciences, Science Technology and Human Values, Perspectives on Science, Mind, the Journal of Philosophy, the Monist, and the Journal of the History of Ideas. The four general philosophy journals are included because the specialist journals did not yet exist at the start of the window: Philosophy of Science began in 1934 and the British Journal for the Philosophy of Science in 1950. Retrieval is by an OpenAlex title-field query, deliberately restricted to titles because OpenAlex abstract coverage collapses before about 1970 and a title-and-abstract query would manufacture the very increase the study sets out to measure. The window is divided into ten true decade bins. Within each bin, candidate records are ranked by citation count at a single snapshot date and the top five are selected, so that measured change cannot be an artifact of the literature growing. Only works published in English in their original appearance are eligible, because hedging and stance in a translated text are properties of the translator. The unit of analysis is the whole article, with every frequency reported as a rate per ten thousand words alongside the raw count. A separate and clearly declared purposive supplement of up to fifteen canonical books is coded under a within-book span rule fixed before any book is opened, reported apart from the journal sample and never pooled with it. Reliability is single-coder with a blind twenty percent recode reported as intra-rater agreement, stratified by frame. The protocol also states what it supersedes. An earlier draft of the manuscript carried a twenty-seven-work coding sheet and a forty-two-record bibliographic file, both drawn from the reference list of that draft rather than from any population, and an appendix that reported genuine query strings while explicitly declining to report hit counts. None of those can serve as a sampling frame, and this protocol makes hit counts and per-stage screening counts mandatory. A feasibility measurement was run before the protocol was written, in order to establish that the frame is populated in every decade bin, including the first. Counts only, no selection. The script that performed it is included in this deposit. Frozen version: 1.0SHA-256 of sampling_protocol_5_1.md:cab36fad1226cfeebd6611f3d73fe7fbeac1c099c88901000ed028813964c145 Any change after this deposit is published as a dated amendment in a new version under the same concept DOI, and any amendment made after coding has begun is reported in the manuscript's limitations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.