Government AI social scoring with detrimental treatment

Remove any general social or trustworthiness score built from behaviour or personal traits from government decision paths; decide on criteria tied to the specific program.

In government-context services (benefits eligibility, permits, inspections, enforcement prioritisation), the decision logic and any model or LLM prompt behind it use only inputs relevant to that specific decision, never a cross-context score aggregated from behaviour history, social media, associations or inferred personal traits. Delete or stop computing social_score / trustworthiness_rating style fields, and do not ask a model to rate a person's general worth or reliability. Where a risk model is legitimately used for one program, document its inputs and keep its output from being reused in unrelated contexts to deny, sanction or deprioritise people.

Where it goes: 1 application source code, 2 data models, 7 prompt construction, 9 AI output handling.

What reviewers look for: no field, model output or prompt that rates a person's general social behaviour, trustworthiness or worth (social_score, trustworthiness_rating, categorical_valuation) feeding a denial, sanction or lower priority; decision functions whose inputs are program-specific criteria with a stated reason on each adverse outcome; scores from one program not read by another program's adverse-action path.

score = social_score(behaviour_history(person), inferred_traits(person))
if score < 40:
    application.deny(reason='low citizen score')

elig = housing_eligibility(income=app.household_income, size=app.household_size,
                           residency=app.residency_status)   # program criteria only
if not elig.ok:
    application.deny(reason=elig.reason)   # reason tied to a stated criterion

Engineering guidance, not legal advice. Each provision below may add its own details (a cadence, a deadline, a required notice element): open it for those.