Atlas / Free reports / Legal

Political Classification as Expert Evidence

What survives cross-examination, what does not, and the boundary a classification instrument must not cross.

An asserted characterisation has nowhere to go under pressure

Proceedings that turn on the political character of an organisation — asylum and country guidance, extradition, proscription challenges, defamation — usually turn on a characterisation nobody has had to justify systematically.

The vulnerability is structural rather than factual. An expert who states that an organisation is extremist, or is not, has offered a conclusion with no visible derivation. Opposing counsel does not need to disprove the conclusion. They need only establish that no method produced it.

The practical difficulty is that country experts are slow to instruct and often unavailable inside the timetable, and many available experts are regional specialists with visible political commitments that opposing counsel will find.

Three properties that hold up

The method predates the instruction. A codebook published, dated and version-controlled on a public website before the case existed cannot have been built to reach its conclusion, and can be shown not to have been. This atlas carries a public revision log with thirty scores changed on evidence, each with its reason recorded.

The limits are structural, not conceded. Confidence ratings and contested flags are part of the instrument rather than admissions made under pressure. An expert who has already published what the instrument cannot do is in a materially stronger position than one who concedes it in the box.

There is a record of the method producing unwelcome answers. A check added after it found ten errors verifies that the codebook’s own named example cases sit where it says they do. The Pirate Parties had been cited as the cultural midpoint while scoring −07. All ten were corrected and the check now runs continuously.

Worked positions

Organisations that recur in proceedings, coded on all six axes with evidence attached to each cell and a declared confidence rating.

Note the confidence column. Two of these are coded at the lowest rating in the dataset, because independent documentation is thin — which is a fact a report should carry rather than smooth over.

What distance does and does not establish

A comparison between two organisations returns a single figure and a composition. The figure is a headline; the composition is the story, and it is the composition that is admissible reasoning.

Two organisations can sit close together for entirely different reasons. The composition column is what an argument is actually about, and it is the column a report should lead with.

The limit of what is offered

Stating the boundary is not a weakness. An expert who claims less than the instrument supports is harder to dismantle than one who claims more, and a report that declines a question outside its method is more credible on the questions inside it.

There is a second boundary worth marking. Whether a movement is fascist, or extremist, or authoritarian, is frequently a dispute about which features define the term. Two people with identical coordinates in front of them can disagree indefinitely and both be reasoning correctly. No coordinate system resolves a definitional dispute; what it does is narrow the argument to what the word should mean, which is a smaller contribution and a real one.

The weakness to know before instructing

This atlas has one coder. It therefore cannot report inter-coder reliability — the standard evidence that a coding scheme produces the same result in different hands, which the Chapel Hill surveys, the Manifesto Project and V-Dem all publish.

That is named on the method page as the instrument’s principal limitation, and it would be raised in cross-examination whether or not it appeared in the report. It appears in the report.

Two things stand in its place, and both are weaker than the real figure. The anchor-case check described above, which tests whether the codebook describes its own data. And a blind coding instrument that shows any reader a cell’s evidence with the score withheld and asks them to assign one — which is the only honest way to establish whether the codebook is precise enough to work in another person’s hands.

Practical points on instruction

• Confirm the procedural requirements of the jurisdiction for expert declarations before instruction. Requirements for form, declaration wording and duties to the court vary.
• A preliminary opinion is usually the cheaper first step. It establishes whether the evidence assists and how far it can be pushed, before a full report is commissioned.
• The methodology annex is annexed, not summarised. A summary invites the question of what was left out.
• Conference and hearing attendance are quoted separately at a day rate.

The other three reportsMore for this role