Method · Context
Argupedia, fact checks and voting advice tools
Three tools for political orientation, three different questions. A fact check tests whether a claim is true. A voting advice tool shows which party shares a position. Argupedia evaluates what consequences a measure has and how they weigh against each other. None replaces the others; where one ends, the next begins.
Whether a claim is true
Fact-checking teams such as PolitiFact, FactCheck.org, Full Fact or, in Germany, Correctiv test individual statements from speeches, posts and programmes against evidence and rate them: true, false, misleading, unsupported. The procedure is fast, traceable and effective against invented numbers and twisted quotes.
Its limit lies in the kind of question. A fact check tests facts, and political debates consist mostly of forecasts: a speed limit will prevent accidents, rent control will slow down construction. Whether a forecast comes true cannot be tested today, only how well it is supported. And a true claim says nothing yet about its weight: that a speed limit costs travel time is true; how heavily that weighs against the accidents avoided is not a question of fact.
Argupedia starts where the fact check ends. The plausibility of an argument is scored under fixed evidence rules: which kind of evidence carries it, how cleanly the cause is identified, and which ceiling that sets. A fact check is one source among others.
Which party shares a position
Voting advice applications such as the German Wahl-O-Mat, run by the Federal Agency for Civic Education since 2002, or iSideWith in the United States present a set of statements. In the Wahl-O-Mat the parties answer 38 statements themselves with agree, disagree or neutral; users answer likewise and receive their agreement with each party in percent. The tools are widely used, transparent in their procedure and make positions comparable that are otherwise scattered across manifestos of a hundred pages.
They answer a question about positions, not about consequences. Whether agreeing with a statement is a good decision stays open: a voting advice tool measures proximity, not effect. Two people with the same position on a speed limit can think completely differently about its consequences, and a party can agree to a statement whose implementation would harm its own goals.
Argupedia does not evaluate parties but proposals. Where a proposal comes from a party manifesto, the party is named as its author, and the election pages (in German) show per party the balance of its scored proposals, derived from the individual evaluations, never as an overall grade. The question “who is close to me” stays with the voting advice tool; the question “what does the proposal I am close to bring about” begins here.
Three tools side by side
| Fact check | Voting advice tool | Argupedia | |
|---|---|---|---|
| The question | Is this claim true? | Which party shares this position? | What consequences does this measure have, and how do they weigh against each other? |
| The object | A single factual claim, mostly about the present or the past | A statement and the parties' answers to it | A concrete proposal with its pro and con arguments over the period of its effect |
| The result | True, false, misleading or unsupported | Agreement in percent per party | Probability that the measure comes out ahead, size of the effect, share of positive effect |
| The values | None needed: a fact is being tested | Your own, through your answers to the statements | Published in the baseline and the value catalogue, the same for every debate, swappable in the Werte-Shift |
| The source of the position | Evidence gathered by the editorial team | The parties answer the statements themselves | Studies, statistics and precedents, with source and calculation path per argument |
| Updating | With new evidence, mostly as a new article | Rebuilt for each election | Continuously, with versions and a review log per measure |
How consequences weigh against each other
Every measure on Argupedia carries its pro and con arguments, and every argument gets a weight from three factors: which value is affected, how large the effect is, and how plausibly it is supported. The three factors stand at every argument with reasoning, sources and calculation path. The balance adds up the weights of both sides and says how likely the measure is to come out ahead.
Weighing needs values, and here lies the difference to every silent trade-off: the values are published. The Baseline (in German) names the factor by which life years, tonnes, hours and euros come onto one scale; the value catalogue names which dimension an argument prices and with what weight. Both apply to every debate alike. Whoever weighs differently recomputes in the Werte-Shift. Why every evaluation needs such a setting is explained on the page about the Münchhausen trilemma.
This tool has limits too, and they are stated on every page. The first evaluation is produced by an AI following the procedure; every page names which model reviewed it last and whether a human has read it. The settings in the baseline are reasoned, not proven. And the balance is a probability, not a judgement on the position someone holds on the measure. Argupedia is neither a fact check nor a voting recommendation. It is the step in between: the consequences of a proposal, weighed in the open, with a yardstick every reader can check.
Going deeper
- Questions about the method: the most frequent questions and every change to the method, with dates.
- The Münchhausen trilemma: why every evaluation ends at a setting.
- Baseline (in German): the conversion anchors that bring consequences onto one scale.
- Data: the whole stock for download, for research and private use.