Morphos — English worked example

English worked example

A dossier, measured.

The 2026 UK Labour Party leadership crisis and succession: Keir Starmer's resignation as Labour leader and Prime Minister, Andy Burnham's Makerfield by-election and leadership bid, and the reactions of party factions, unions, opposition parties, media and the public.

The finding in one sentence

The media and the public didn’t frame it the same way: Political / strategy was a media frame, the media carried it 20 points more than the citizen-reachable layer — while Morality ran the other way, 10 points stronger in the citizen-reachable layer.

traces to 45 coded documentsframe facet α 0.57 — tentative9.2% of the corpus, coded sample

This is a prototype, produced on a dossier we did not pick to please you. The version for your dossiers — your actors, your window, your reading — is in preparation.

What this demonstration answers, in order

Which frames dominate this dossier? Who introduced each one — and who are its loudest relays? Where do you sit in that landscape?

Act I — the lineage

Who set the frame, and who followed — relay by relay.

The genealogy of the dossier’s most-inherited frame: who introduced it, then each sphere that picked it up, dated, with the lag against the originator. Institutional relays link to the original document; the citizen relay is metadata only (never named). The story the map then proves.

morphos · lineagereal data

11 Mayoriginthe union floor
Dave Ward (CWU general secretary) and CWU conference delegates Unions
The « Policy » frame enters the corpus from the union floor — its first documented carrier.
d_b691235ab09a3740 d_8382a95f1156322b d_b64f1c39228dea5d
14 May+3 dgovernment
Keir Starmer / Prime Minister's Office State
Government picks up the same frame, 3 days after the originator.
d_f74dbfbfc9150627
18 May+7 dthe press
westerntelegraph.co.uk Media
The press picks up the same frame, 7 days after the originator.
d_bcdf8fecfa213e50 d_4fa549b3e528b0fb d_3a1c058252e96bcb
29 May+18 dexpert commentary
Tom Fleming Experts
Expert commentary picks up the same frame, 18 days after the originator.
d_2129ba860c08037d d_719faacbe2f1432a d_6154e06107b1a998
1 Jun+21 dparty politics
Zia Yusuf (Reform UK Shadow Home Secretary) Parties
Party politics picks up the same frame, 21 days after the originator.
d_37d1308b3da0e80e d_33f54e3d81aa997d d_cc302daa5c367404
22 Jun+42 dthe public
a citizen source Citizens
The public picks up the same frame, 42 days after the originator.
d_87dfb0a66b122a56 d_02b31d295da7d572
One frame, 6 spheres — from the union floor to the public, in 42 days.
The genealogy is a measurement prose cannot produce: who introduced the frame, who inherited it, and in what order. Every relay above is dated to its first documented carrier in the coded subset.

lineage

Enable JavaScript to walk the genealogy.

Act II — the actor map

Who is in the discourse, and who they reference.

morphos · actor mapreal data

Period cumulative by week

Links

actor map

Enable JavaScript to explore the interactive map.

Every circle is an actor layer or a named public figure/institution, sized by how much it is named; its colour is its camp; arrows are who references or quotes whom. Named subjects sit with the actor layer that most references them — who keeps each subject in the story. The citizen layer is an aggregate (the peach node): individuals are never named or linked (analysis-only rail). Positions are deterministic; every number is from the coded sample.

Act III — diffusion

Which frames spread, and when.

Weekly share of voice of the top 8 frames across 2026-04-15 → 2026-07-17 — the largest is the top band. Drag the cursor or the chart; the readout gives the weekly shares. Every value is from the coded subset (9.2% of the corpus). The bands are frames, so the reliability that applies is the frame facet’s: tentative (α 0.57, below the 0.80 bar): read the direction, not the decimal.

morphos · diffusionreal data

Political / strategy, Policy, Economic

Enable JavaScript to explore the weekly diffusion.

Act IV — the gap

Media and the public didn’t frame it the same way.

For each frame, its share in media against its share in the citizen-reachable layer (public networks + unions). Political / strategy is the widest gap; the flagged row is the most evenly shared — the common ground. Click a bar for sample documents (institutional linked to the original, citizens counted only, never named). Shares are within the coded subset, and these are frames, so the reliability that applies is the frame facet’s: tentative (α 0.57, below the 0.80 bar): read the direction, not the decimal.

morphos · the gapreal data

Political / strategy

Enable JavaScript to explore the gap frame by frame.

Method & honesty

What these numbers can — and cannot — say.

Coding coverage

9.2%926 of 10,061 sentences coded, across 45 of 328 documents (13.7%). Whole documents, coded completely, stratified by week × actor type — 45/47 strata. 2 strata dropped (2026-05-25 · Citizens, 2026-06-01 · Citizens) Every engine number describes this coded subset, not the whole corpus.

Reliability per facet (α) · 0.80 bar

Frame

0.57tentative (below bar)

Narrative

0.82at or above bar

Theme

0.79tentative (below bar)

Rhetoric

0.75tentative (below bar)

Krippendorff α vs the human gold slice (uk, en, 60 sentences) — disclosure, not a gate. Tentative facets are machine first-pass, not adjudicated. Every view on this page is built from the frame facet, so that is the row its numbers answer to; the others are disclosed because the coder measures them, not because a figure here rests on them.

Coder: machine_luna_v9 (gpt-5.6-luna, prompt coder-prompt-v9-20260717, effort high). Reproducible: the same corpus and codebook re-run to the same numbers. No frame-level finding is presented as solid, and no share is presented as the whole corpus.

The vocabulary, defined

Six coded facets, used in one sense throughout. They are separate measurements with separate reliabilities: a figure computed on one facet never carries another facet's reliability.

frame: The value lens a sentence uses — how an issue is made to matter. Drawn ONLY from one frozen list of 15 tokens shared by every dossier (Economic, Policy, Political, Fairness, Legality, Capacity…), which is what makes frames comparable from one dossier to the next. A sentence may carry several, or none.
narrative: A named, contested story about who did what, entered in THIS dossier's registry; each narrative hangs off exactly one parent frame. Narratives are dossier-specific and are coined by the analysis; frames are not. A ranking of narratives and a ranking of frames are different measurements.
theme: The subject matter a sentence is ABOUT — emergent labels, not a controlled list. Orthogonal to frame, and never collapsed into it: two sentences can share a theme and mobilise opposite frames.
rhetoric (rhetorical device): The device the sentence uses to say it — metaphor, hyperbole, irony, loaded diction, a coined label, and so on. HOW it is phrased, never what it claims. The inventory is on trial and may be narrowed; an empty cell is a real value.
stance: The position a sentence takes toward a NAMED narrative: it advances it, contests it, or reports it. Always narrative-scoped — one sentence can advance one narrative and contest another, which is the signal the gap is built from.
impact (field of influence): The documented traces of transmission between actors: dated, evidence-bearing edges, the adoption of a signature phrase, and who carried a frame first. It is a record of what travelled, never a claim about what caused what — see the ladder below. Reach and audience response are not measured at all.

What the evidence can support — strongest first

Every statement about transmission sits on one of these rungs, and says which. The top rung is listed to be refused, not used.

demonstrated causal — not available from this data X caused Y. NEVER claimable from this data: nothing here measures outcomes, audiences or counterfactuals, so this rung exists only to be refused.
documented transmission An evidenced, dated edge: this actor quoted, referenced or took up that actor's wording. The strongest thing the data supports.
temporal precedence X used it first, Y followed. That is ORDER, not influence — the first documented carrier is not “the origin”.
association Both are present in the corpus with no ordering established.
inferred A reading of the pattern, carried as a reading and marked as such.