The five failure modes
How a one-dimensional model produces false statements, in five recognisable ways.
Five ways, not infinite ways
A one-dimensional model does not fail randomly. It fails in five recognisable ways, each with a signature. Four of the five are usually committed in good faith by people who are not trying to deceive anyone.
1 · Substitution
Establish something true about axis A, then state a conclusion about axis B, and omit any argument connecting them. The most common failure in political argument by a wide margin.
What makes it persistent is that step one is usually correct. Challenging it produces more evidence for step one, because that is the part the speaker can defend — which reads to them as your refusing to accept a well-supported fact.
| Organisation | EC | CU | AU | SC | ID | CH |
|---|---|---|---|---|---|---|
| National Socialism | +02 | +09 | +10 | +09 | +10 | +09 |
| Classical Marxism | −10 | −06 | +02 | +04 | −09 | +08 |
| gap | 12 | 15 | 8 | 5 | 19 | 1 |
The claim that the Nazis were socialists. Twelve points apart on ownership — the axis the claim is about. Much closer on the organisation of power, scale and method. Right that a resemblance exists; wrong about where it lives.
2 · Collapse
An error of representation rather than inference: it happens in the act of compressing six numbers into one. The mapping is many-to-one and not reversible, so an enormous number of genuinely different positions produce the identical score, and there is no way even in principle to recover which one you started with.
This is how “the extremes meet” gets manufactured. A line that records distance from the centre while discarding direction reports the surviving quantity as a shared position.
3 · Drift
Two quantities get confused. Absolute position is where an organisation sits on a fixed scale. Relative position is where it sits with respect to the current distribution.
An organisation can appear to move while standing still, when the midpoint moves. It can appear consistent while moving a long way, when it moves with the midpoint.
Ordinary political language reports relative position almost exclusively. Arguments about whether an organisation has changed are about absolute position — so both parties produce evidence and neither can settle it.
4 · The name trap
A name is evidence of exactly one thing: what an organisation wanted to be called when it chose the name.
The instructive case is not the deceptive one. The Liberal Party of Australia is the principal conservative party of its country and its name is entirely honest — it uses liberal in the nineteenth-century sense. The trap does not require anyone to lie. Names are fixed at a moment and vocabulary keeps moving.
The replacement rule: read what an organisation legislated, expropriated, banned and built. Four categories, in rough order of evidential weight, all checkable by anyone with a library card.
5 · Asymmetry
The selective application of scrutiny — examining one side’s claims for the four failures above and accepting the other side’s at face value.
Why this one is invisible from inside
It does not feel like bias. It feels like rigour. You are applying a demanding standard, every error you catch is real, and the process is honest at every individual step. The bias is entirely in the sampling, and sampling is not visible from within the sample — you cannot notice the claims you did not examine.
The external symptom: if the method has never yet produced a result you disliked, you are not running it.
Learning this material makes the failure more attractive, not less. Before, you had a vague sense that political labels are sloppy. Now you have five named modes and worked examples, which is a considerably better toolkit for demonstrating that your opponents are confused. The toolkit works. That is the problem.
The symmetry test
Ten minutes, in writing.
- Write the claim your own side makes most confidently about its opponents. Not a caricature — the version you would defend in public. One sentence.
- Name the axis it asserts.
- Ask what evidence establishes it on that axis, not an adjacent one.
- Ask whether it relies on what the opponent calls themselves, or on what they legislated, expropriated, banned and built.
- Check whether it is about absolute or relative position.
- Now run the identical procedure on the strongest claim the other side makes about you.
Step six is the whole exercise. Steps one to five are comfortable. The expected result is that your claim is partly right: accurate on one axis, overstated in scope. That is the characteristic finding, and there is no reason to expect yours to be the exception.
One or two emails a month, or none at all
Nine free modules, nothing gated, nothing to fill in. The update carries codings that changed, new claim checks, and at least one finding that went against the atlas. Nothing else, and one click to stop.
Site owner: set FORMSPREE_ID in offers.py to activate this form.