Skip to content

romantic_relationships.yaml

Controls sexual orientation assignment and romantic partnership formation. Two modes are available: data-driven (ONS-derived per-area probabilities, recommended for UK worlds) and YAML-only fallback (hand-tuned national probabilities, used when area-level data is unavailable).

Topic: Relationships
Path: configs/2021/relationships/romantic_relationships.yaml

Keys

Key Description
name Arbitrary label for this relationship config
min_age / max_age Age bounds for orientation and partnership assignment
data_sources Paths to ONS-derived prevalence and per-area CSV files; omit to use YAML fallback
diagnostics Logging options for verification runs
sexual_orientations Orientation types, probabilities, age adjustments, and compatibility rules
storage Keys written to person.properties for orientation and relationship status

name, min_age, max_age

name: "romantic_relationships"
min_age: 16
max_age: 120

name is used only for logging. min_age and max_age define the age window for assignment; people outside this range are skipped entirely for both orientation and partnership steps.

data_sources

data_sources:
  prevalence_path: "data/population/sexual_orientation/orientation_prevalence_extended.csv"
  msoa_marginal_path: "data/population/sexual_orientation/orientation_by_msoa_normalized.csv"
  geo_level: "MGU"

When present, the engine uses raked probabilities: national prevalence by sex × age is combined with per-area orientation marginals to produce P(orientation | sex, age, area). prevalence_path points to the national-level file; msoa_marginal_path to the per-area file; geo_level names the geography level at which area codes in that file are expressed.

Omit this entire block for worlds without area-level orientation data — the engine falls back to the sexual_orientations.probabilities values below.

diagnostics

diagnostics:
  verbose: false

verbose: true enables detailed per-run logging: a national-vs-empirical orientation comparison, MSOA quintile sweep, and assignment diagnostics. This walks the full population multiple times — leave false for production runs, enable only when verifying a code change.

sexual_orientations

sexual_orientations:
  types:
    - heterosexual
    - homosexual
    - bisexual
  probabilities:
    male:
      heterosexual: 0.90
      homosexual: 0.05
      bisexual: 0.05
  age_adjustments:
    "16-25":
      bisexual: 1.3
    "75+":
      homosexual: 0.7
  compatibility:
    heterosexual:
      male: [female]
      female: [male]
    bisexual:
      male: [male, female]
      female: [male, female]

types lists the orientation labels in the order they will be processed. Any string is valid; the labels must be consistent with probabilities, age_adjustments, and compatibility.

probabilities provides fallback national-level probabilities keyed by sex then orientation. Used when data_sources is absent or when an area is missing from the marginals CSV. Values need not sum to 1.0 — the engine re-normalises after any age adjustments.

age_adjustments applies multiplicative adjustments to the base probabilities for a given age band. Keys are age-range strings in the format "min-max" or "min+". Each adjustment is applied as a multiplier to the named orientation's probability before re-normalisation. This allows, for example, higher bisexual rates among younger cohorts or lower rates for specific orientations in older age groups.

compatibility maps each orientation and sex to the list of partner sexes that are considered compatible. The engine uses this when searching for a partner: a candidate is eligible only if their sex appears in the compatibility list for the person's orientation and sex. Any sex values defined in the population may be listed; there is no restriction to male/female.

storage

storage:
  orientation_key: "sexual_orientation"
  status_key: "relationship_status"

orientation_key is the key written to person.properties for the assigned orientation string. status_key is the key written for partnership status. Both keys must also be listed in serialization_config.yaml under population.properties to appear in the HDF5 output.