Replication package: Measuring Otherness Between Political Communities — survey responses, codebook and analysis script

Responses of two applications of an otherness perception instrument, de-identified, with a codebook and a single script that recomputes every figure reported in the accompanying article and report. The instrument measures otherness on three axes — feeling, relation and status — and asks the same thirty items of both sides of a pair, so that each community's reading of the other is recorded separately. Every item is scored from −20 to +20. The first application was fielded in August and September 2024 and measured citizens of Turkey against Syrian immigrants resident in Turkey. It was administered in two languages, Turkish and Arabic. After exclusions, 210 respondents: 190 on the Turkish side, 20 on the Arabic side. The second application was fielded in July 2025 and measured Turks against Kurds, using twenty-nine of the same thirty items. After exclusions, 613 respondents: 398 Turkish, 169 Kurdish, 45 belonging to neither community, 1 unassigned. Both are non-probability convenience samples recruited online. They support no claim about the communities concerned; they are evidence about the instrument. The script requires no libraries beyond the Python standard library. It prints each figure beside the value printed in the article and marks the line OK or DIFFERS: running python3 analysis.py reports 60 checks, 60 matching the article. What is published. Both surveys were administered anonymously. Neither collected a direct identifier: no name, e-mail address, telephone number, place of residence or birth, IP address or device fingerprint. Every closed-ended answer is published exactly as recorded — categories are not merged, small cells are not collapsed, and no subsample has its demographics withheld. What is withheld. Free text is not published: response identifiers, timestamps, the home-town field, the open description of discrimination experienced, the description of one's personal relationship with the other community, and the platform comment field. Written-in text is the only place where a respondent could have named an employer, a role, a place or an incident that would allow a reader to work out who he is. Where a respondent typed into an "other, please specify" box instead of choosing from the list, the answer appears as other. README.md sets out the procedure and its limits in full. Files. study2025.csv and pilot2024.csv hold the responses; codebook.md defines every column and gives the thirty items in full; analysis.py computes the reported figures; expected_output.txt records what it prints; prepare_data.py and build_pilot.py show how the two response files were produced from the raw exports. Version notes — 3.0.0 Two changes, both of them corrections to what the earlier releases withheld. No item response and no reported figure has changed; the analysis script still reports 60 checks, 60 matching the article. Closed-ended questions that were asked but not deposited have been restored. Their omission was an oversight in the scripts that built the files, not a decision, and it was not noticed because the published analysis did not use them. Restored to the second application: the question on which value most needs protecting (God, State, Individual, Society), social media use, the platform used most, and Christian denomination. Restored to the first application: school type, social media platform, whether political views are shared online, best foreign language, mother tongue, the same value question, membership of a religious group, membership of a political group, having been subjected to discrimination, and how often the respondent encounters racist or exclusionary situations. The value question was asked in both applications, which makes it comparable across them — something the earlier releases made impossible. The suppression applied in earlier releases has been reversed. Those releases merged every category held by fewer than five respondents, collapsed the Alevi denomination, and withheld the demographics of the two small subsamples — the twenty Syrian respondents of the first application and the forty-five respondents of the second belonging to neither community. That was a precaution rather than a finding. Faith and denomination are central to a study of otherness, and withholding them cost the record more than it protected. Absent evidence that some combination of these closed answers identifies a particular person, the answers now stay as recorded. Related works İlişki Kayıt Is supplement to Measuring Otherness Between Political Communities: Decomposition, Bidirectionality and Distance(makale — DOI çıkınca eklenecek) Is supplement to Aradaki Mesafe: Türkler ve Kürtler Ötekilik Algısı Araştırması 2025 (Vicdan Vakfı raporu) Version notes — 2.0.0 2.0.0. Rebuilt the first application from its two original form exports, one in Turkish and one in Arabic; the file used for version 1.0.0 had every "no knowledge" and "not interested" response blank rather than scored, so the escape responses never entered the computation. Fourteen records excluded, none on anything answered to the thirty items: test submissions made while the second platform was being built, and respondents who gave an age band reaching below eighteen. Sections 7.3 and 8.1 of the article record both changes.