Method · Edition 01
How a society's culture is turned into a number
Everything behind the scores, in the order it is calculated. Plain enough to check, precise enough to rebuild, and split into steps you can disagree with one at a time.
01
The thesis
Culture is the sediment of change. Politics moves in months, ideology in years, culture in generations. A model that reads only the top layer is fitting the weather and calling it climate.
The practical consequence: culture is a slow state variable that sets the bounds of what is politically feasible. It is not a sentiment indicator. It changes slowly, it changes unevenly across cohorts, and when politics leaves it behind, the gap itself becomes predictive.
02
Three clocks, one frame
Every reading in this app belongs to exactly one layer. The layers differ in how fast they move, which is why they cannot be averaged together.
| Layer | Clock | What it contains |
|---|---|---|
| Culture | generations | Values, belonging, trust, cohesion. The sediment that sets what a society will accept. |
| Ideology | years | Coalitions, issue salience, media diet. How cultural material gets sorted into camps. |
| Politics | months | Elections, approval, legislation. The visible weather above the sediment. |
03
Six dimensions of culture
Culture is not one thing, so the bottom layer is read along six separate dimensions. Each is scored 0 to 100 on the same stated direction, so a higher number always means more of the named condition.
Identity & Belonging
identityLanguage of us and them, flags, faith, migration sentiment across social and media streams.
Institutional Trust
trustConfidence voiced in government, courts, media and experts; measured as the inverse of institutional blame.
Economic Mood
economicFelt security: rent, wages, work. Higher means more secure, not richer.
Openness to Change
opennessTolerance for social change, immigration, technology and unfamiliar norms.
Social Cohesion
cohesionInter-group trust, civic participation, street-level civility. The inverse of polarisation.
Generational Alignment
generationalHow far the youngest cohort's values sit from the oldest. Higher means generations still agree.
The culture score is the weighted mean of the six. No dimension is allowed to dominate: a society cannot buy a high culture score with one strong reading.
culture_score = mean(identity, trust, economic, openness, cohesion, generational)
Reported to one decimal place. Twelve-month change is the same mean computed against readings from a year earlier.
04
Four evidence streams
Each stream is trusted differently. A hashtag is evidence of talk; a protest, a full stadium or a budget line is evidence of behaviour. Durability is the weight applied to a reading before it enters a dimension.
| Stream | Horizon | Weight | What it yields |
|---|---|---|---|
| Social platformsX, TikTok, Instagram | days to weeks | 1.0× | Volume, velocity and framing of everyday cultural talk. |
| Media coverageNational press, broadcast, newsletters | weeks to months | 1.6× | What a society is being told matters, and in which register. |
| City eventsProtests, festivals, attendance, incidents | months | 2.4× | Behaviour in public space, the hardest evidence of belief. |
| Policy spendingBudgets for culture, education, welfare, defence | years | 3.2× | Where a state puts its money, before it puts its words. |
05
From signal to score
A reading enters the model as an intensity between 0 and 100, weighted by the durability of its stream and decayed by age. The decay half-life is set per stream, because a viral week and a spending commitment do not expire at the same rate.
dimension = Σ (intensity × durability × decay) / Σ (durability × decay)
decay = 0.5 ^ (age_days / half_life_days)
half_life: social 42d · media 120d · city 270d · policy 730d
Normalised to 0–100 within a rolling 24-month window. Readings older than the window drop out; a dimension with no fresh evidence holds its prior reading and is flagged low-confidence.
Two guards keep the composite honest. A dimension needs at least three independent readings to move, and no single stream may contribute more than 60 per cent of a dimension score. When either guard trips, the dimension is reported as provisional rather than silently averaged.
06
The strain model
The point of the index is not the ranking. It is the distance between the layers. Strain is politics minus culture: how far the fast layer has travelled beyond the slow one.
strain = politics_score − culture_score
projected_politics = politics_score − (strain × closure_rate)
Closure rate is the share of the gap a society has historically closed within twelve months, set per society from its past contests.
Positive strain means politics is ahead of its cultural base. History offers three resolutions: correction at the next contest, realignment of the coalition, or a period of volatility that resolves neither. Negative strain means culture is pulling politics forward, and the next cycle usually moves toward the cultural position.
Treat the projection as a prior, not a forecast. It says which direction pressure points and how strong it is, not who wins. For a team already producing numbers, the strain reading enters as an additional prior that constrains the outer bounds of an outcome distribution, rather than shifting its centre.
07
Rules for reading the index
- Read the strain, not the level. A high culture score is not a prediction on its own.
- Check the drift before the number. A falling culture score under rising pressure matters more than either fact alone.
- Find the carrying dimension. Strain concentrated in one dimension resolves faster than strain spread across all six.
- Compare within the region. Absolute scores across continents mostly measure survey access.
- Never read a single stream. Social data leads; behaviour confirms; spending entrenches.
08
What this edition is not
Election results, turnout and platform reach are published figures, taken from national electoral commissions and the platforms' own advertising tools via DataReportal. The culture, ideology and politics scores are the model's reading of that material, not a survey result, so treat them as an index rather than a measurement. Platform years before the latest are back-cast estimates and marked as such.
Four biases would survive even a fully wired version, and the model has to carry them openly:
- Platform access bias: what is measurable on X, TikTok and Instagram is shaped by who is allowed to post, and by which APIs still exist.
- Language and literacy skew: English and Spanish dominate machine-readable discourse, which over-weights urban, younger and online populations.
- Amplification noise: coordinated and bot activity inflates volume, which is why intensity is capped and behaviour streams carry the heaviest weight.
- Policy lag: budget lines are commitments, not outcomes, and they arrive a year or more after the culture that produced them.
The next step is ingestion: connect the four streams to live sources, hold a human review queue before any reading is marked verified, then back-test the strain model against past elections in the region.
