Data

One compact JSON file with four parts. Field names and the controlled vocabulary are German throughout, also in the English debates; the most common ones are explained here.

grundlagen

Foundations: scoring (the constants of Scoring v1.3: value register, conversion anchors, weights, evidence ceilings), werte (the value catalogue with classes and relations) and baseline (the anchors with derivation, range and change history). The prose, in German: Baseline, value catalogue, plausibility.

bereiche[]

Areas: per area the lead, the situation text, indicators and sources, plus massnahmen[] (measures): title, measure text, classification and, per scoring country, the highest version under bewertungen[].bewertung. English debates carry sprache: "en".

bewertung

The evaluation of a measure: arguments[] with side (pro or con), value v, impact i and plausibility p with their three reasonings (begruendung_v/i/p), the calculation path (rechenweg), the value assignment (wert) and source references; bilanz with the probability of a positive net effect (p_positiv), the net effect (netto) and the positive share (anteil_positiv); quellen[] (sources); pruefungen[] (review log). The method behind the fields is explained in the framework essay and the FAQ.

wahlen[]

The German state elections 2026 with their proposals from party manifestos, in the same form as the areas; each proposal names its policy area in bereich_slug.

Older evaluation versions are not in the file; every debate page names its versions, and the full history is available on request.

The dataset is licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). Use, reproduction, adaptation and redistribution are permitted for research, teaching, non-commercial journalism and private purposes, under three conditions: attribution to “Argupedia, argupedia.com” with a link to this page, a reference to the licence, and clear marking of your own changes to the data.

The evaluations, texts and the procedure remain protected by copyright; the licence grants rights of use and transfers no ownership. Commercial use, including in products, models or services with a commercial purpose, only by agreement: info@argupedia.com.

The sources an evaluation refers to are not part of the dataset; they belong to their respective authors and are listed in quellen[] with title and address.

Every evaluation of a debate rests on values, and every justification stops somewhere at a setting. Argupedia names those settings and keeps them the same for every debate. The file makes that checkable: whoever weighs differently can recompute, and whoever holds a number to be wrong can say exactly which one. The FAQ answers the questions this raises.