Jersey States Election · June 2026 · Data Dossier

The Bio Dossier

Ninety-two candidates. Roughly 220,000 words of self-promotion. One question: what are they actually saying? Below: rhetorical positioning, sentiment, AI-suspicion, clichés, ego indexes and constituency overlap — all derived from the public bios on vote.je.

1 · The Fiscal × Social Quadrant

Two views of the same chart. Bio rhetoric shows where each candidate's bio language sits. Researched record overrides bios with positions derived from public record (Hansard, party platforms, ministerial briefs, local press) for the 34 prominent candidates we researched manually. Both, with arrows shows the gap.

Bio rhetoric tells you how candidates present themselves. Researched record tells you where they actually sit. The arrow view is gold for an article: every arrow is a candidate whose self-presentation diverges from their public record.

1b · Where Rhetoric Diverges from Record

Top ten candidates whose bio language sits furthest from their researched political position. Only candidates in the prominent set (party-affiliated + sitting Senatorial candidates) have researched scores; for everyone else this gap is undefined.

The 'says one thing, does another' leaderboard

2 · Sentiment Leaderboard

Average per-sentence sentiment (VADER). Positive = sunny, future-facing, optimistic copy; negative = doom, frustration, "broken Jersey" rhetoric. Most political bios skew positive by definition — campaigners selling hope, not despair.

☀️ Most Optimistic Bios

Top 10 by mean sentence sentiment

🌧️ Grittiest / Most Negative Bios

Bottom 10 — these candidates are not here to comfort you

Sentiment distribution

3 · The AI-Suspicion Leaderboard

A composite score (0–100) for "did this read like ChatGPT wrote it?" — based on em-dash density, telltale phrases ('delve', 'tapestry', 'multifaceted'), uniform sentence lengths, tricolons, lack of personal anecdote, and lexical predictability. It is a heuristic, not a verdict. Some humans naturally write like LLMs; some LLMs are now better at sounding human.

🤖 Most ChatGPT-flavoured Bios

High score = many AI tells, sparse personal detail

👤 Most Human-Sounding Bios

Anecdotes, oddities, typos and all

By Party — average AI-suspicion

4 · The Vocabulary of an Election

Top 30 distinguishing words and phrases across all 92 bios (TF-IDF, with party/parish names and generic election-speak filtered out). What people actually talked about — minus the obligatory "I am standing as a candidate".

5 · Top Election Issues

Bios scored against thirteen issue lexicons. Numbers show total mentions and the share of candidates who raised each issue at least once.

Headline: Housing/cost-of-living and the economy dominate; environment and crime barely register. Government reform and transparency comes up repeatedly but usually as a buzz-phrase rather than a costed proposal.

6 · Cliché Bingo

A small but choice tray of stock political phrases. Tally across all bios. Tick five in a row at your local hustings to win.

7 · The Constituency Overlap Heatmap

Inside any given constituency, are candidates saying genuinely different things — or basically writing the same bio? Cosine similarity (0 = unrelated, 1 = identical) between bios within each constituency. Pick one to inspect.

8 · Per-Candidate Explorer

All 92 candidates, every score. Click a column header to sort. Type to filter.